Argus: Programming with Communication Protocols in a Belief-Desire-Intention Architecture

Reference: Samuel H. Christie V, Munindar P. Singh & Amit K. Chopra (2025). Argus: Programming with communication protocols in a belief-desire-intention architecture. Artificial Intelligence (AIJ), Volume 348, article 104398. Received 26 June 2024; revised 23 July 2025; accepted 24 July 2025. North Carolina State University & Lancaster University. Open access (CC BY 4.0). DOI: 10.1016/j.artint.2025.104398. URL. Code and examples: https://gitlab.com/masr/bspl.

Summary

Protocols model multiagent systems by capturing the communications between agents, while Belief-Desire-Intention (BDI) architectures organise an agent in terms of cognitive concepts — beliefs, goals, and plans. The two traditions are complementary: the agent-oriented tradition (exemplified by Jason) captures the internal reasoning that drives an agent’s interactions, whereas the interaction-oriented tradition (exemplified by information protocols such as BSPL) constrains that reasoning for the sake of interoperability. The authors argue that no prior approach combines them cohesively, and that current BDI approaches lack adequate support for engineering protocol-based agents. Argus — named for the builder of the Argo, the ship that transports Jason on his quest — synthesises the two. It leaves the BDI reasoning engine and the core abstractions of beliefs, goals, and plans unchanged, and replaces only the agent-communication part of a BDI architecture with an information protocol. Agents can thus be authored and executed much as before, with streamlined communication.

Concretely, a multiagent system in Argus is enacted over a protocol specified in BSPL, where each message schema declares a sender role, a receiver role, and parameters adorned ⌜in⌝ / ⌜out⌝ / ⌜nil⌝ with some designated keys. These adornments capture causality and integrity: a parameter that is ⌜in⌝ must already be known to the sender, an ⌜out⌝ parameter is generated by the sender and becomes known, and a ⌜nil⌝ parameter must remain unknown. Because ordering and exclusion fall out of information flow rather than control-flow operators, a compliant agent may send and receive messages in any relative order without ordered or reliable delivery. Each agent keeps a local state — beliefs corresponding to the messages it has observed (sent or received) — separate from its internal state of private beliefs. A generic adapter applies the protocol specification to validate every incoming and outgoing message and to update the local state, which is reflected to the BDI reasoner as ordinary beliefs. Emission plans, triggered by the addition of a belief for an observed message, give the developer a place to plug in the internal reasoning that produces the bindings for a message’s ⌜out⌝ parameters.

Argus contributes (1) a novel architecture and formal operational semantics combining protocols and BDI; (2) a code-generation-based programming model that produces Jason (AgentSpeak) scaffolding capturing changes to an agent’s view of the protocol state and guiding the implementation; and (3) integrity checking (an “Argus Checker”) for incoming and outgoing messages that helps ensure agents are well-behaved. The design makes weak assumptions about infrastructure and preserves both autonomy (an agent may ignore or respond to messages as it sees fit) and heterogeneity (local and internal states are separated, avoiding hidden dependencies), so the approach is readily adapted to other cognitive agent-programming languages such as 2APL, 3APL, CAN, and GOAL. The paper diagnoses concrete limitations in Jason’s communication — no support for protocols, hard-coded KQML performatives and semantics, and selection / social-acceptability functions that are neither public nor grounded in beliefs and plans — and shows the Argus conceptual architecture building naturally on top of Jason’s reasoning cycle. The Argus architecture rests on the Local State Transfer (LoST) style, an information-flow analogue of REST. Rather than prove soundness and completeness theorems, the authors give an informal argument relative to Jason’s reasoning cycle and defer the formal notions to future work. A running “Grading” example — a US-university setting with Professor, Student, and TA roles exchanging beginTest, challenge, rubric, response, and result messages — illustrates the flexibility that information protocols afford.

Key Ideas

Meld two traditions. Keep the BDI engine and the beliefs / goals / plans abstractions unchanged; swap out only the communication layer, replacing it with an information protocol. Jason is the exemplar, but the approach is meant to port to 2APL, 3APL, CAN, and GOAL.
Local state vs internal state. The local state holds beliefs about observed (sent or received) messages and is used to validate emissions and receptions; the internal state holds the agent’s private reasoning. The separation is what preserves autonomy and heterogeneity.
Generic protocol adapter. A reusable adapter validates each incoming and outgoing message against the BSPL specification and updates the local state, which is surfaced to the BDI reasoner as ordinary beliefs; Emitter and Receiver components move messages, and an Argus Checker enforces integrity.
Emission plans. A plan is triggered by the addition of a belief for an observed message and specifies the protocol state the enactment should be in for a further message to be emitted; such a plan is necessarily incomplete, offering the developer a hook to supply the internal reasoning that binds the message’s ⌜out⌝ parameters.
Code generation. A bundled generator emits Jason / AgentSpeak code that tracks the agent’s view of the protocol state and guides the implementation, reducing ad hoc handling and easing the notoriously hard task of testing BDI programs.
Integrity checking. Both incoming and outgoing messages are validated at the adapter so agents stay well-behaved; attempts to emit protocol-violating messages are currently detected and abandoned (a limitation the authors note — a better design would signal the failure to the agent’s own reasoning).
Weak infrastructure assumptions. Because BSPL constrains only emission and reception can be treated idempotently, an unordered, unreliable transport suffices; the choice is justified via the end-to-end argument. No predefined performatives and no dependence on KQML.
Critique of Jason’s communication. Jason adopts KQML primitives and semantics, and uses selection (S_M) and social-acceptability functions to control communication that are neither public nor grounded in beliefs and plans, with inadequate support for autonomy and heterogeneity — the shortcomings Argus sets out to fix.
Local State Transfer (LoST). The architectural style underlying BSPL, compared to REST for web services but going beyond it: LoST promotes decentralised, peer-to-peer enactment where each agent maintains its own local state, rather than REST’s client-server view.

Connections

BSPL
Information Protocols
Interaction Protocols
BSPL - The Blindingly Simple Protocol Language: the protocol language Argus enacts
Semantics and Verification of Information-Based Protocols: the formal semantics of BSPL that Argus builds on
Kiko - Programming Agents to Enact Interaction Protocols: the immediate predecessor programming model (decision makers over forms); Argus lifts the same protocol-adapter idea into a BDI architecture
BDI
BDI Architecture
AgentSpeak: the language whose Jason implementation is Argus’s BDI exemplar
Modeling Rational Agents within a BDI-Architecture: the Rao & Georgeff BDI foundation
JaCaMo
CArtAgO
An Interaction-oriented Agent Framework for Open Environments: a JaCaMo/commitment framework Argus contrasts with
KQML: the performative model Jason inherits and Argus replaces
FIPA-ACL
Commitment-based Semantics
Commitment Machines - Yolum and Singh
Session Types: an ordering-based, globally-specified protocol model the paper critiques
Speech Acts
Agent-Oriented Programming
Multi-Agent Systems
Heterogeneity Resolution
End-to-End Arguments in System Design: invoked to justify enacting protocols over simple, unreliable transports
Local State Transfer
CBCL - Safe Self-Extending Agent Communication

Conceptual Contribution

Claim: BDI agent programming and information protocols are complementary and can be unified without disturbing the BDI reasoning engine — replace only the communication layer of a BDI architecture with an information protocol and you gain flexible, loosely-coupled, verifiable interaction while retaining the cognitive abstractions of beliefs, goals, and plans. Doing so yields more flexible multiagent systems than current BDI practice allows, while preserving agent autonomy and heterogeneity.
Mechanism: A generic protocol adapter that validates incoming and outgoing messages against a BSPL specification and reflects the agent’s observed-message local state as beliefs to the BDI reasoner; emission plans in which developers supply the internal reasoning that binds a message’s ⌜out⌝ parameters; a code generator that produces Jason / AgentSpeak scaffolding for the protocol state; message-integrity checking; and an operational semantics layered on Jason’s reasoning cycle (with soundness and completeness argued informally rather than proved).
Concepts introduced/used: Information Protocols, BSPL, BDI, AgentSpeak, Local State Transfer, Interaction Protocols, Heterogeneity Resolution
Stance: architecture + programming model / engineering, with a formal operational semantics
Relates to: Extends the BSPL programme — BSPL - The Blindingly Simple Protocol Language, Semantics and Verification of Information-Based Protocols, and especially Kiko - Programming Agents to Enact Interaction Protocols — from bespoke decision-maker agents to mainstream BDI / Jason agents. It contrasts with commitment-in-BDI work such as Commitment Machines - Yolum and Singh and JaCaMo, and with ordering-based global models like Session Types. The “decide from local history, validate against the protocol” shape it shares with Kiko - Programming Agents to Enact Interaction Protocols is the same one that underlies CBCL - Safe Self-Extending Agent Communication’s trace verifier.

Tags

#bdi #agent-programming #information-protocols #bspl #jason #interaction-protocols #operational-semantics #aij

Backlinks

Linked Pages

CBCL - Safe Self-Extending Agent Communication

CBCL: Safe Self-Extending Agent Communication

Reference: O’Connor, H. (2026). “CBCL: Safe Self-Extending Agent Communication.” 2026 IEEE Symposium on Security and Privacy Workshops (SPW), pp. 39–48. Preprint: arXiv:2604.14512v1 [cs.CR], 16 Apr 2026. Source file: oconnor-cbcl-langsec26.pdf. URL. Reference implementation: https://codeberg.org/anuna/cbcl-rs.

Summary

Building on McCarthy’s 1975/1982 Common Business Communication Language vision and the language-theoretic security programme of Sassaman, Patterson, Bratus and Locasto (Security Applications Of Formal Language Theory), this paper presents a contemporary CBCL: an agent communication language designed so that all messages — including runtime language extensions — remain within the deterministic context-free (DCFL) complexity class. The motivating diagnosis is that contemporary agent communication has slid up the Chomsky hierarchy without acknowledging the security cost: KQML and FIPA-ACL are CFL with informal extensibility, but MCP and LLM-based agent frameworks operate over inputs whose effective complexity is recursively-enumerable, where the validity of an arbitrary message is undecidable and “weird machines” (Exploit Programming - From Buffer Overflows To Weird Machines) become unavoidable. CBCL takes the opposite tack: minimal core (8 performatives — tell, ask, reply, hello, bye, ok, error, cancel) plus a homoiconic dialect mechanism that lets agents define, transmit, and adopt domain-specific vocabularies as first-class CBCL messages, with three machine-checked safety invariants — R1 (no recursion in dialect templates), R2 (declared resource bounds on depth, expansion size, verification time), R3 (core-vocabulary preservation) — that together guarantee DCFL preservation under arbitrary finite sequences of dialect installations.

The paper accompanies the design with a Lean 4 formalization (≈5,400 lines, 16 modules, 176 machine-checked theorems, zero sorry gaps, no custom axioms) covering parser correctness (soundness + completeness), R1–R3 verification, pipeline totality, and the central dcfl_preserved closure theorem; a verified parser binary cbcl-parse is extracted from the proofs. A separately-developed Rust reference implementation (cbcl-rs, ≈11,000 lines, no_std-compatible core) ships a hand-rolled O(n) parser, dialect verification engine, gossip-based dialect propagation (O(log n) convergence per Demers et al.), C FFI/WASM bindings, property-based and differential tests, and cargo-fuzz targets; differential tests cross-validate the Rust parser against the Lean-extracted binary, eliminating spec-implementation parser differentials. Five example dialects (precision agriculture, AI planning, cross-chain asset transfer, artifact management, email) are R1–R3-verified end-to-end. A draft IETF Internet-Draft specifying application/cbcl and a Nostr transport binding (cbcl-nostr) accompany the paper. The Discussion explicitly aligns the design with Singh’s social-agency programme: dialect installation is treated as a public commitment to the dialect’s declared semantics — semantic divergence becomes a breach of commitment, observable and accountable, rather than a hidden failure of shared mental state — sidestepping the unverifiable mentalistic semantics critique of Verifiable Semantics for ACLs.

Implementation status (post-paper, cbcl-rs). The reference implementation has shipped two further safety invariants the paper sketches as future work, plus a richer Lean formalisation: R4 (Integrity) — Ed25519-style signatures over canonical encodings via an algorithm-agnostic Signer trait, with three-valued install verdicts (Valid / Unsigned / Invalid) — and R5 (Contract Well-formedness) — optional (protocol …) and (shape …) clauses on a dialect, expressing causal Merkle DAG protocols and per-performative shape contracts; R5 enforces acyclicity, reachability from begin, performative definedness with ancestor closure for extends, step uniqueness, and depth-bound respect, all decidable in linear fuel at install time. The §VII-D structural-contracts vision is therefore live (SPEC-002, SPEC-003): verify_monotone, verify_all_is_meet, and verify_eventually_consistent are proved theorems showing causal-protocol verification is a lattice homomorphism that joins coordination-free under G-Set store merge — explicit CALM Theorem alignment. The Lean codebase has grown to 380+ declarations across 19 files (still zero sorry, only the standard axioms propext / Classical.choice / Quot.sound); the workspace ships 837 tests including 584 cbcl-core unit tests, 37 proptest cases, 21 differential integration tests against the Lean-extracted binary, libFuzzer harnesses, and cargo-mutants mutation testing at a 90% kill threshold. The crate layout has grown from the paper’s five to seven: the original cbcl-core / cbcl-parser / cbcl-cli / cbcl-wasm / cbcl-ffi plus cbcl-erl (Erlang/LFE binding, SPEC-009 — with SPEC-010 binding-conformance suite) and cbcl-arena (Agent Arena platform, SPEC-012 — composable referee on cbcl-lfe-router with capability-keyed message routing and signed receipt logs). A worked sealed-bid auction dialect (SPEC-004) demonstrates structural defence against false-claim manipulation — concrete evidence that the R5 contract layer is rich enough to encode mechanism-design integrity properties decidably. The earlier Scheme prototype (anuna-research/cbcl) is superseded; architecture decisions live as ADRs under .hence/ (purity boundary, parser library choice, error handling, WASM API surface, R4 adversarial review).

Key Ideas

DCFL as the Goldilocks complexity class: Regular is too weak for nested envelopes/dialects; general CFL admits ambiguous grammars and parser-differential attacks; context-sensitive is PSPACE-complete; Type-0 is undecidable. DCFL is the minimal class that supports nested structure with class-level unambiguity (parser equivalence under language evolution).
Homoiconic self-extension: Dialect definitions are themselves valid CBCL messages, parsed by the same recogniser used for ordinary communication — a single attack surface, no separate meta-language. Inspired by Racket’s #lang mechanism.
Three safety invariants R1–R3: declarative pattern-template substitution only (no recursion, no iteration, no reflection); declared resource bounds (depth ≤ 64, expansion ≤ 8192 chars, verification ≤ 1000 ms); the eight core performatives cannot be redefined.
DCFL preservation theorem: dialect invocations must appear inside a (lang name …) wrapper; the lang tag plus a unique dialect name forms a prefix-free partition that lets a deterministic pushdown automaton dispatch to the appropriate subgrammar without extra lookahead — circumventing the fact that DCFLs are not closed under union in general. Mechanised in Lean 4 as dcfl_preserved.
Lean 4 formalization with extraction: 176 theorems, 0 sorry gaps, no custom axioms; the verified parser binary cbcl-parse is extracted from the proofs and runs the same fuel-bounded recursive-descent parser as the model.
Rust reference implementation: cbcl-core (no_std, no unsafe), cbcl-parser, cbcl-cli, cbcl-ffi, cbcl-wasm. Differential tests against the Lean-extracted binary; cargo-fuzz over five entry points.
Eight core performatives (tell, ask, reply, hello, bye, ok, error, cancel) plus four message categories (simple, meta, lang-scoped, wrapped); keyword parameters in Common-Lisp style (:thread, :in-reply-to).
Single-pass template expansion: terminates in time linear in the template plus argument size; not Turing-complete; output is not fed back into the expander, ruling out unbounded re-expansion.
Epidemic dialect propagation: gossip-based with Demers et al.’s O(log n) convergence bound; agents independently verify R1–R3 and signatures before installing.
Canonical serialization per RFC 9804 for cryptographic operations — deterministic byte representation as a precondition for signature verification.
Dialect identity by content hash: name collisions resolved by canonical-form hash; downgrade attacks rejected at installation.
Self-expression vs self-adaptation (Zambonelli et al.): traditional ACLs support only parameter tuning within fixed structure; CBCL is a self-expressive protocol whose vocabulary structurally evolves under safety constraints.
Singh-aligned reading of dialect installation: a public commitment to declared semantics, with :examples clauses providing machine-checkable behavioural anchors — semantic divergence as observable breach rather than hidden mental-state mismatch.
Stated limitations: DCFL ceiling means no recursive transformations, no cross-field references (e.g. checksums over fields), no aggregation over variable-length collections, no context-sensitive validation. Standish-taxonomy paraphrastic only — no Orthophrase, no Metaphrase. Trust infrastructure (key distribution, revocation) and dialect-curation policy left to deployment.
Structural-contracts extension (sketched in §VII-D, shipped post-paper as R4 + R5): Ed25519-style dialect signatures plus protocol constraints as causal Merkle DAGs over performative-typed messages, with Visibly Pushdown Languages for shape constraints and coordination-free monotonic verification over an append-only store. Lean-mechanised lattice-homomorphism theorems (verify_monotone, verify_all_is_meet, verify_eventually_consistent) establish that replica verdicts join under G-Set merge.

Connections

Common Business Communication Language — McCarthy’s 1975/1982 origin; CBCL is the formally-verified realisation of that vision.
Security Applications Of Formal Language Theory — Sassaman et al.’s LangSec foundations.
Proof-Carrying Code - Necula — Necula 1997, the architectural ancestor of CBCL’s R4 split (host runs a small certified verifier; producer ships the proof).
Enforceable Security Policies - Schneider — Schneider 2000, the formal warrant for R5 monotone-verdict trace monitoring.
An Axiomatic Basis for Computer Programming - Hoare — Hoare 1969, root of the contract-based verification line CBCL inherits.
Authentication in Distributed Systems - Lampson Abadi Burrows Wobber — Lampson et al. 1992, the Speaks-For Calculus for delegated commitments.
Articulating Reasons - Brandom — Brandom 2000, the philosophical underwriting of Public Semantics (entry point).
Making It Explicit - Brandom — Brandom 1994, the full inferentialist treatise.
Empiricism and the Philosophy of Mind - Sellars — Sellars 1956, the Myth of the Given that makes Semantics-Without-Minds possible.
Philosophical Investigations - Wittgenstein — Wittgenstein 1953, Meaning As Use / language-games / rule-following.
Two Dogmas of Empiricism - Quine — Quine 1951, Semantic Holism.
Assertion - Stalnaker — Stalnaker 1978, Common Ground update semantics — operational analogue of CBCL’s verdict ledger.
Languages and Language - Lewis — Lewis 1975, Convention of Truthfulness picture of how an abstract language gets in force in a population.
Communicative Actions for Artificial Agents - Cohen Levesque — Cohen & Levesque 1995, the Mentalistic Semantics high-water mark CBCL deliberately is not.
FIPA-ACL Specifications — the engineering of FIPA-ACL (envelope, conversation-id, ontology slot) CBCL inherits; the semantics it replaces.
An Ontology for Commitments in Multiagent Systems - Singh — Singh 1999, the formal commitment ontology CBCL’s dialect contracts instantiate.
Jones-Sergot Normative Systems — Jones & Sergot 1993, the Counts-As Relation that gives wire events institutional meaning under a dialect contract.
Exploit Programming - From Buffer Overflows To Weird Machines — weird-machine analysis CBCL is designed to preclude at the parser layer.
PKI Layer Cake - Kaminsky Patterson Sassaman — concrete parser-differential attacks; CBCL’s class-level unambiguity rules these out at the grammar level.
The Halting Problems of Network Stack Insecurity
A Language-Based Approach To Prevent DDoS
LangSec
KQML as an Agent Communication Language
KQML - A Language And Protocol For Knowledge And Information Exchange
FIPA-ACL
ACL Rethinking Principles — Singh’s social-agency critique that CBCL’s “dialect installation as public commitment” inherits.
Agent Communication Languages - Rethinking the Principles
Verifiable Semantics for ACLs
A Common Ontology Of ACLs
Commitment-based Semantics
BSPL: Singh’s Information Protocols (AAMAS 2011) are the nearest multiagent-systems precedent for the R5 causal protocols and the R6 role layer: declarative, asynchronous, decentralised enactment from local history with no ordering assumptions. CBCL adds content-addressed predecessors, an adversarial threat model with signature-bound roles, the DCFL bound, runtime dialect installation, and a mechanised local-to-global correspondence; Enactability is the BSPL-side notion nearest to causal locality.
The State of the Art in Agent Communication Languages
Trends in Agent Communication Language
Toward Principles for the Design of Ontologies Used for Knowledge Sharing — Gruber’s ontological-commitment principle CBCL inherits.
Some Philosophical Problems from the Standpoint of Artificial Intelligence — McCarthy & Hayes’s representation/computation separation.
Elephant 2000 - A Programming Language Based on Speech Acts — McCarthy’s intra-program speech-act counterpart to CBCL’s inter-organisational vision.
Adjectival Modifiers
S-expression
Homoiconicity
Code as Data
Creating Languages in Racket — Flatt’s #lang mechanism that inspired CBCL’s homoiconic self-extension.
Extensibility in Programming Language Design - Standish
Language Extensibility Taxonomy
Paraphrase
Orthophrase
Metaphrase
Self-Expression
Self-Adaptation Self-Expression Self-Awareness ASCENS
Dialects and Idiolects
Keeping CALM - When Distributed Consistency is Easy — CALM theorem underwriting the forthcoming coordination-free contracts extension.
Visibly Pushdown Languages
MCP Landscape Security Threats And Future Research Directions
Survey Of AI Agent Protocols
Survey Of Agent Interoperability Protocols
ClawWorm Self-Propagating Attacks Across LLM Agent Ecosystems — concrete LLM-agent worm whose root-cause analysis CBCL’s lang-scoped provenance addresses.
Why AI Agents Communicate In Human Language
Why Do Multi-Agent LLM Systems Fail
Gossiping in Distributed Systems
Gossip-based Aggregation in Large Dynamic Networks

Conceptual Contribution

Claim: Agent communication languages can be simultaneously expressive, runtime-extensible, and tractably verifiable iff every message — including those that extend the language — is constrained to the deterministic context-free language class; this constraint is compatible with first-class self-extension, can be machine-checked end-to-end, and converts the unverifiable-mental-state problem of mentalistic ACLs into a public-commitment problem at the protocol layer.
Mechanism: Eight-performative core grammar (LL(1), DCFL); homoiconic dialect mechanism — dialect definitions are valid CBCL messages with extend clauses describing pattern-template substitutions; three safety invariants R1–R3 enforced at installation (no recursion via DFS cycle detection, declared resource bounds, core-vocabulary preservation); (lang name …) wrapper providing prefix-free deterministic dispatch into per-dialect sub-grammars so the union remains DCFL. Lean 4 formalization (LeanCbcl, ~5,400 LoC, 176 theorems) covering parser soundness/completeness, R1–R3, pipeline totality, and dcfl_preserved; verified parser extracted to a runnable binary (cbcl-parse). Rust reference implementation (cbcl-rs, ~11,000 LoC) with a ResourceContext-tracked single-pass expander, gossip-based dialect propagation (O(log n) per Demers et al.), canonical RFC-9804 serialization for signing, algorithm-agnostic Signer trait, C FFI / WASM bindings, property-based + differential tests, fuzz targets. Draft IETF application/cbcl Internet-Draft and Nostr transport binding.
Concepts introduced/used: Deterministic Context-Free Language, LangSec, Homoiconicity, S-expression, Dialects and Idiolects, Performatives, Self-Expression, Paraphrase, Orthophrase, Metaphrase, Visibly Pushdown Languages, CALM Theorem, Commitment-based Semantics, Public Semantics, Verifiable Semantics, Adjectival Modifiers, Ontological Commitment, Lean 4, Property-Based Testing, Parser Differential Attack, Weird Machine, Gossip Protocols, Canonical Serialization, Nostr, hence
Stance: engineering / formal-methods, with explicit position-taking on the agent-communication design space
Relates to: Realises McCarthy’s Common Business Communication Language vision (1975/1982) by giving formal safety guarantees McCarthy could only gesture at. Sits squarely in the LangSec programme of Security Applications Of Formal Language Theory and Exploit Programming - From Buffer Overflows To Weird Machines, extending it from input-validation analysis to protocol design — claiming, plausibly, to be the first ACL designed from LangSec principles and the first to demonstrate that LangSec is compatible with runtime self-extension. Inherits the DCFL-vs-ambiguous-CFG argument from PKI Layer Cake - Kaminsky Patterson Sassaman and applies it at the message layer rather than the certificate layer. Builds explicitly on Singh’s social-agency critique (ACL Rethinking Principles) and Wooldridge’s verifiability programme (Verifiable Semantics for ACLs) by treating dialect installation as a public commitment whose semantics are anchored by the :examples clause — a machine-checkable analogue of the commitment-trace conformance that Flexible Protocol Specification and Execution (Yolum & Singh) and Commitment Machines - Yolum and Singh establish for protocol actions. Where Pact - A Choreographic Language for Agentic Ecosystems takes the individual-rationality fork to answer “why follow a protocol?”, CBCL takes the publicness/verifiability fork: the meaning of a message is fully determined by the formal grammar plus declared semantics, with no mental-state inference required. Inherits Gruber’s ontological-commitment architecture but replaces “out-of-band agreement” with in-protocol propose/query/teach. The extensibility design draws on Creating Languages in Racket (Flatt’s #lang) and respects the boundary set by Extensibility in Programming Language Design - Standish: paraphrastic only, deliberately excluding orthophrase/metaphrase to keep the attack surface bounded. The forthcoming structural-contracts extension uses Visibly Pushdown Languages for shape constraints and the CALM Theorem for coordination-free verification — a natural bridge to commitment-machine semantics. The LLM-agent threat-model framing (ClawWorm Self-Propagating Attacks Across LLM Agent Ecosystems, MCP Landscape Security Threats And Future Research Directions) positions CBCL as the structural alternative to MCP/JSON-RPC-with-natural-language: same extensibility, vastly tighter security envelope.

Kiko - Programming Agents to Enact Interaction Protocols

Kiko: Programming Agents to Enact Interaction Protocols

Reference: Samuel H. Christie V, Munindar P. Singh & Amit K. Chopra (2023). Proc. 22nd International Conference on Autonomous Agents and Multiagent Systems (AAMAS 2023), London, May 29 – June 2, 2023. IFAAMAS, 10 pages. Posted as arXiv:2606.26156v1 [cs.MA], 23 June 2026. URL. Software and appendix: https://gitlab.com/masr/bspl/-/tree/kiko

Summary

Kiko is a protocol-based programming model for agents that enact BSPL information protocols. Its diagnosis is that existing agent programming offers poor abstractions for decision making: JADE hard-wires FIPA’s ordering-based protocols, Jason and JaCaMo give cognitive abstractions but no protocols, commitment-based approaches lean on centralised commitment stores or do not operationalise asynchrony, and agent-oriented methodologies leave protocols as informal UML diagrams. Kiko’s answer is that the messages an agent sends are its public decisions, so the programmer’s job should reduce to writing decision makers: functions that are handed the set of messages the agent is currently enabled to send and choose which of them to complete and emit.

The enabling abstraction is the form: a prototype message instance whose ⌜in⌝ parameters are already bound from the agent’s history (its causal dependencies are met) and whose ⌜out⌝ parameters are left for the decision maker to fill. A generic protocol adapter computes the enabled forms from the agent’s local store, invokes decision makers on events, and validates each returned set of instances as an emission attempt: either every instance in the set is emitted or none is, so a decision maker that tries to send both Buy and Reject for one enactment is rejected outright rather than partially honoured. The adapter also validates receptions for integrity, and because BSPL constrains only emission, no ordering, reliability or central coordination is required of the transport; the implementation runs over UDP and treats reception as idempotent. Worked patterns include correlation across concurrent enactments, cross-enactment decisions (buy the cheapest of several quotes), playing roles in several unrelated protocols at once, atomic emission sets spanning protocols, reception-order freedom (a Rescind may overtake the Quote it depends on and simply disables the matching Buy), loose coupling under protocol change, and single-form decision makers as a convenient special case.

The formal part models an agent as a history plus input and output channels and gives a transition semantics with three rules: Recv (receive a message not already in the history if it passes the receive-check), Tx (delivery, whose non-application models loss) and Decide (compute enabled forms, apply a decision maker, emit if the send-check passes). Theorem 1 (soundness) states that every reachable state of a Kiko multiagent system simulates some reachable enactment of the protocol, so compliance holds regardless of what the decision makers choose; Theorem 2 shows that an optimised Decide rule that checks only the internal compatibility of the chosen set is equivalent to the full check, since forms are drawn from the enabled set and decision makers preserve their bindings; Theorem 3 (completeness) states that for any reachable enactment there exist decision makers that realise it, assuming all sent messages are received. The discussion relates Kiko to its predecessors in the same programme (Stellar, which introduced forms but used per-message handlers; Protocols over Things; Bungie; Mandrake), invokes the end-to-end argument for building over simple transports, and sketches norm-based forms, accountability, concurrent decision makers as actors, and microservices as directions.

Key Ideas

Messages are decisions. A protocol constrains decision making, not message ordering; an agent’s programming interface should be the set of valid decisions currently open to it.
Forms. Enabled message prototypes with ⌜in⌝ bound and ⌜out⌝ unbound, derived incrementally per enactment context (MAS identifier plus key bindings) from the local store.
Decision makers. Programmer-written functions from sets of forms to sets of instances; invoked on communication events by default, or on custom events via an internal queue, so no polling.
Atomic emission sets. An attempt is emitted in full or rejected in full; partial emission of a consistent subset is ruled out as arbitrary. This is the feature the authors call unique to Kiko.
Adapter architecture. Local Store, Enablement, Checker, Emitter and Receiver; the adapter is generic and understands information protocols, so the communication service is fully abstracted from business logic.
Unordered, lossy transport suffices. BSPL constrains only emission and reception is idempotent, so UDP without ordering or reliability enacts protocols correctly; this is offered as compatibility with the end-to-end argument, and as something ordering-based protocol approaches cannot claim.
Formal definitions. Association, instance, form, context, consistency, out- and nil-compatibility, derivation, enablement, decision maker, send-check and receive-check (Definitions 2 to 13).
Soundness, optimisation, completeness. Reachable states simulate protocol enactments (Thm 1); the internal-compatibility-only Decide is equivalent (Thm 2); any reachable enactment is realisable by some decision makers under delivery (Thm 3).
Loose coupling. Because coupling is by information, adding an indirect bank-transfer path to Purchase leaves the seller’s decision logic unchanged; the adapter derives the Deliver form once payment arrives by either route.
Lineage. Stellar (forms, message handlers), Protocols over Things and Mandrake (fault tolerance), Bungie (extensible application-level protocols), Tosca and Clouseau (commitments compiled to information protocols).

Connections

BSPL
Information Protocols
Enactability
BSPL - The Blindingly Simple Protocol Language
Semantics and Verification of Information-Based Protocols: defines the reachable enactments that Kiko’s states are shown to simulate
Interaction Protocols
Agent-Oriented Programming
AgentSpeak
An Interaction-oriented Agent Framework for Open Environments: the JaCaMo-based commitment framework Kiko contrasts with
Agents and Artifacts
FIPA-ACL Specifications: the JADE/FIPA ordering-based incumbent
Commitment-based Semantics
Actor Model: proposed substrate for concurrent decision makers
Multiparty Asynchronous Session Types: cited as a representative ordering-based protocol language
Endpoint Projection
CALM Theorem
CBCL - Safe Self-Extending Agent Communication

Conceptual Contribution

Claim: Agents should be programmed as sets of decision makers over the messages a protocol currently enables them to send. Given an information protocol, this abstraction guarantees compliance by construction, supports realistic decision patterns that ordering-based models cannot express, and needs only an unordered, unreliable transport.
Mechanism: Forms derived from the local store per enactment context; decision makers returning emission attempts checked atomically; a protocol adapter (store, enablement, checker, emitter, receiver) over UDP; an operational semantics (Recv, Tx, Decide) with soundness, an equivalent optimised check, and completeness relative to BSPL’s reachable enactments.
Concepts introduced/used: Information Protocols, BSPL, Enactability, Agent-Oriented Programming, Actor Model, Interaction Protocols
Stance: programming model / engineering, with formal guarantees
Relates to: Completes the BSPL stack begun in BSPL - The Blindingly Simple Protocol Language and Semantics and Verification of Information-Based Protocols by giving the agent side a programming interface whose correctness is stated against that semantics. It occupies the place that per-role projected programs occupy in Choreographic Programming and per-role monitors occupy in Session Types, but reaches it without Endpoint Projection: the adapter evaluates the global protocol against the agent’s own history. The same “decide from local history, verify from the trace” shape appears in CBCL - Safe Self-Extending Agent Communication’s R5 verifier, which is why Kiko and its BSPL lineage are the nearest multiagent-systems precedent for that work.

Session Types

A type discipline for communicating processes — originating with Honda (1993) and developed extensively by Honda, Vasconcelos, Yoshida, and others — in which a process’s communication behaviour is typed by a session type: a structured description of the sequence of messages that may flow on a channel, including alternation (!/?), choice (+/&), recursion, and delegation. Session types come in two flavours: binary (two participants per channel) and multiparty (a global type projected to per-role local types — directly analogous to choreographic Endpoint Projection). When a process can be source-checked, its session-type conformance is decidable statically; when it cannot, runtime monitors can enforce the type from observable message traces (On the Monitorability of Session Types). Structurally adjacent to the commitment-based protocol tradition: session types and Yolum-Singh-style commitment protocols are independently-developed formalisms for typing communicating agents against a publicly-observable global protocol.

In this vault

Multiparty Asynchronous Session Types — Honda, Yoshida & Carbone POPL 2008, the multiparty session-types foundation
Structured Communication-Centred Programming for Web Services — Carbone, Honda & Yoshida 2007/2012, choreographies built on session types
Introduction to Choreographies
On the Monitorability of Session Types
Choreographic Programming
Endpoint Projection
Commitment Machines - Yolum and Singh
BSPL: the multiagent-systems counterpart that dispenses with projection and ordered channels; Enactability in place of projectability
Flexible Protocol Specification and Execution
Conversation Protocols
Pact - A Choreographic Language for Agentic Ecosystems

Commitment Machines - Yolum and Singh

Commitment Machines

Reference: Yolum, P. & Singh, M.P. (2002). Intelligent Agents VIII: Agent Theories, Architectures, and Languages (ATAL-01), LNAI 2333, pp. 235–247. Springer. (A preliminary version appeared in WET-ICE 2000.) Source file: yolum-singh-2002-commitment-machines.pdf. URL

Summary

Yolum and Singh propose commitment machines (CMs) as a protocol formalism that, unlike finite-state machines or Petri nets, attaches declarative meaning to states and actions in terms of the participants’ Social Commitments. A CM is a triple ⟨M, Δ, F⟩ of a finite minimal set of meanings (formulas in a propositional language extended with a commitment operator Cₓp and a strict-implication operator ;), a set of actions ⟨a : e⟩ whose effect e is itself a meaning, and a set of consistent final meanings. Transitions are not enumerated — they are inferred from the logical consequences of applying an action’s effect to the current meaning. The running example is the NetBill Protocol, specified by atomic propositions (request, goods, pay, receipt) and abbreviations for conditional commitments (accept = C_c(goods ; pay), promiseGoods = C_m(accept ; goods), etc.); three reasoning rules govern how commitments evolve as content propositions are made true.

The paper’s central technical contribution is a compilation procedure that turns any complete CM (one whose meaning set is closed under antecedence) into a deterministic FSM that can be executed in constant space without commitment reasoning at run time. The procedure is shown sound (Theorem 2 — every realised computation is generated by the source CM) and complete (Theorem 3 — every CM-generated computation has a semantically-strongest, efficient FSM-realised counterpart from true), via a chain of lemmas about semantic superiority, efficient computations, and endpoint equivalence. Because the meanings of states are explicit, agents executing the CM directly can plan paths through the protocol that exploit opportunities (e.g. a merchant proactively quoting without a request, “try before you buy” deals) and handle exceptions, recovering autonomy that pure-token FSM/Petri-net specifications eliminate.

Key Ideas

Commitment machine ⟨M, Δ, F⟩: state space is a finite minimal set of meanings (formulas) rather than tokens; actions ⟨a : e⟩ carry effect-meanings; final states are a consistent subset of meanings.
Inferred transitions: q ⊨_⟨a:e⟩ r iff (q ∧ e) ⊢ r — transitions are derived from logical entailment, not enumerated edges.
No fixed start state: a CM has no s₀; participants may begin from any meaning whose commitments they accept (typically true).
Three reasoning rules over commitments: (1) Cₓp discharges when p holds; (2) Cₓ(p ; r) collapses to base-level Cₓr when p holds; (3) Cₓ(p ; r) ceases when r holds before p, with no new commitment.
Compilation to deterministic FSM under two restrictions: (R1) no transition mᵢ → mⱼ if mᵢ ⊢ mⱼ (no information-free moves); (R2) the target meaning must dominate all alternatives entailed by the action (forcing maximal information).
Completeness requires “complete” CMs — meaning set closed under antecedence — so that a unique strongest target meaning always exists; trivially achieved by closing under conjunction.
Soundness, completeness, determinism established as theorems over the computation abstraction (sequences of meanings interleaved with actions); computations are compared by semantic superiority and endpoint equivalence rather than literal action-sequence equality.
Flexibility recovered for free: since states encode meanings, agents can skip steps, reorder, or take unsolicited shortcuts whenever the resulting meaning is still in M and the path to a final meaning still exists.

Connections

Conceptual Contribution

Claim: Multi-agent protocols should be specified by giving states and actions declarative meaning in terms of the participants’ commitments — not by enumerating legal token sequences — and any such specification can be soundly and completely compiled to a deterministic FSM, recovering executability without forfeiting flexibility, autonomy, or the ability to handle exceptions and exploit opportunities.
Mechanism: A propositional language with a commitment operator Cₓ and strict-implication ;; meaning sets M and final-meaning sets F (consistent w.r.t. logical consequence); action effects given as meanings; transitions inferred via (q ∧ e) ⊢ r. Compilation Procedure 1 maps ⟨M, Δ, F⟩ → ⟨S=M, Σ={a}, s₀=true, Q=F, δ⟩ under the two information-monotonicity restrictions; Procedure 2 promotes a computation to its semantically strongest counterpart, Procedure 3 removes redundant self-transitions, yielding efficient semantically-strongest computations that are realised by the compiled FSM. Soundness (Thm 2), determinism (Thm 1), and completeness for complete CMs (Thm 3) are proved.
Concepts introduced/used: Commitment Machines, Commitment-Based Protocol, Commitment-based Semantics, Conditional Commitment, Social Commitment, Public Semantics, Verifiable Semantics, NetBill Protocol, Endpoint Projection (structural analogue), Conversation Protocols, Speech Act Theory
Stance: formal-semantic / engineering
Relates to: Direct precursor and structural sibling of Flexible Protocol Specification and Execution (Yolum & Singh, AAMAS-02), which lifts the same six commitment operations (Create / Discharge / Cancel / Release / Assign / Delegate) into the Event Calculus and adds an abductive planner; ATAL-01 here is the FSM-compilation half of that programme. Companion to A Commitment-Based Approach to Agent Communication (Fornara & Colombetti, AAMAS-02 / AAI-04), which gives an OO/FSM operational semantics for the same primitives plus a precommitment state for directives. Carries forward the social-agency programme of ACL Rethinking Principles (Singh 1998) and answers Wooldridge’s challenge in Verifiable Semantics for ACLs by grounding meaning in publicly-observable commitment formulas rather than agent mental state. The Discussion explicitly contrasts with Smith et al.’s joint-intention–based protocols (mental constructs vs. social constructs) — the same fork that Intention Is Choice with Commitment anchors on the mental side. Structurally a direct analogue of choreographic Endpoint Projection: both compile a globally-specified protocol artefact to a per-role executable, the CM/FSM compilation losing declarative meaning for runtime efficiency much as endpoint projection loses the global view for distributable execution. This makes ATAL-01 the load-bearing reference for the constructive synthesis suggested in the Pact - A Choreographic Language for Agentic Ecosystems entry — choreographies could carry both communication structure (Pact-side) and a CM-style commitment trace (Singh-side), with one compilation step yielding both deadlock-freedom and conformance-checkability.

Tags

AgentSpeak

Rao’s (1996) operational BDI agent programming language: an agent is a triple of belief base (Prolog facts), plan library (event-triggered procedures), and intention stack (plans-in-execution). The interpreter cycles through perceive → event-generate → plan-select → execute. Eight-rule SOS-style operational semantics fits on a page. Made practically dominant through Jason (Bordini-Hübner-Wooldridge 2007), the most-deployed BDI platform in academic MAS research.

In this vault

Semantics and Verification of Information-Based Protocols

Reference: Munindar P. Singh (2012). Proc. 11th International Conference on Autonomous Agents and Multiagent Systems (AAMAS 2012), Valencia, Spain, June 4–8, 2012. IFAAMAS. PDF. Semantic Scholar

Summary

The sequel to BSPL - The Blindingly Simple Protocol Language gives BSPL the formal semantics its flexibility demands, defines three correctness properties, and implements verifiers for them. A protocol denotes a set of enactments over a universe of discourse of roles and message schemas. Each role has a history of the messages it has emitted or received; a history vector collects one per role. Viability is the local rule from the 2011 paper made precise: a role may emit a message only if it knows the bindings of every ⌜in⌝ parameter for the enactment’s key and knows no binding for any ⌜out⌝ or ⌜nil⌝ parameter; reception is always viable and extends knowledge. The intension of a protocol is the set of history vectors that are viable, respect the fundamental causality constraint that a reception is preceded by the corresponding emission, and are covered by the intensions of the protocol’s references.

Three properties are then defined on the intension. A protocol is enactable if its intension is non-empty, that is, some viable path binds all public parameters. It is safe if no history vector binds the same parameter twice for the same key: safety is the distributed integrity property, and the paper’s diagnosis is that across roles it holds precisely when every pair of conflicting ⌜out⌝ emissions is gated by a causally prior branching point controlled by a single role. It is live if from any reachable state the enactment can still legally complete; liveness entails enactability but not the converse. The small examples fix intuitions: Local Conflict is safe but not enactable, Abrupt Cancel (a race between buyer and seller on outcome) is enactable but not safe, Purchase Unsafe drops the private parameter that made accept and reject exclusive, and Purchase No Ship is enactable but not live because after accept nothing can ever bind outcome.

For verification, a protocol’s causal structure is expressed as a set of declarative temporal precedence constraints, and each property is decided by checking the (un)satisfiability of the conjunction of that causal structure with a property-specific formula using a temporal reasoner. The paper reports the clause counts for Purchase and its variants and for the FIPA Request protocol, and lists cheaper design-flaw checks (deadwood messages, parameters that can never be bound or used). The related-work discussion positions BSPL against AUML, message sequence charts and choreography description languages, all of which take a procedural stance on ordering, and notes that both of Desai and Singh’s enactability failures, blindness and nonlocal choice, yield an empty intension and are thereby caught. The LoST middleware (local state transfer) is cited as the architecture that enforces the local constraints at runtime.

Key Ideas

Semantics by enactments. A protocol’s meaning is its intension: the set of viable, causally consistent history vectors covered by its references. Local views and asynchrony are built into the model rather than abstracted away.
Viability is local. Emission requires knowing all ⌜in⌝ bindings and no ⌜out⌝ or ⌜nil⌝ binding for the enactment key; the semantics of a protocol depends only on the specification, never on imagined internal policies of the agents.
Declarative force of ⌜out⌝. An ⌜out⌝ binding is not a report but a declaration (Austin), which is why it may occur only once per key and why a means clause can attach a commitment to it.
Enactability, safety, liveness. Non-empty intension; at most one binding per parameter per key across all roles; completion always remains possible. Liveness entails enactability.
Safety across roles needs a same-role branching point. Two conflicting ⌜out⌝ emissions by different roles are safe only if a causally prior conflict controlled by one role makes them exclusive; the private response parameter in Purchase is the canonical example.
Causal structures. The partial-order structure of a protocol as a set of temporal constraints, mappable to a finite-state machine only with an explosion in states, so reasoning is done declaratively instead.
Verification by satisfiability. Each property is a formula conjoined with the causal structure and handed to a temporal reasoner; validated on Purchase variants and FIPA Request.
Blindness and nonlocal choice both yield empty intensions, so the enactability check subsumes the earlier diagnostics of Desai and Singh.
Design-flaw checks. Deadwood messages, never-bound private parameters, never-used private parameters, protocols with public ⌜in⌝ parameters (empty intension).

Connections

BSPL
Information Protocols
Enactability
BSPL - The Blindingly Simple Protocol Language
Kiko - Programming Agents to Enact Interaction Protocols: the operational semantics whose soundness is stated relative to the reachable enactments defined here
Interaction Protocols
Commitment-based Semantics
Speech Act Theory: the declarative force of ⌜out⌝ bindings
FIPA-ACL Specifications: FIPA Request as a verification benchmark
Knowledge of Choice
Realizability and Verification of MSC Graphs
Deciding Choreography Realizability
Message Sequence Charts - A Survey
Multiparty Compatibility in Communicating Automata
Endpoint Projection
CBCL - Safe Self-Extending Agent Communication

Conceptual Contribution

Claim: An information-based protocol language can be given a formal semantics that respects the locality of each role, the flow of causality across roles and the asynchrony between them, and its central correctness properties (enactability, safety, liveness) can be formulated declaratively and verified mechanically.
Mechanism: Universe of discourse, role histories, history vectors, viability, and intension by cover; Definitions of enactability (non-empty intension), safety (key uniqueness in every history vector) and liveness (completion always possible); causal structure as temporal precedence clauses; verification by (un)satisfiability with a temporal reasoner; evaluation on Purchase variants and FIPA Request.
Concepts introduced/used: Enactability, Information Protocols, BSPL, Realizability, Knowledge of Choice, Interaction Protocols
Stance: formal semantics / verification
Relates to: Enactability is the multiagent-systems analogue of the Realizability problem for message sequence charts and choreographies (Realizability and Verification of MSC Graphs, Deciding Choreography Realizability) and of Knowledge of Choice in Choreographic Programming: all ask whether independently acting endpoints can jointly produce exactly the specified global behaviour. BSPL’s answer is stated over local knowledge rather than channel orderings, which is also the reading that the CBCL role layer’s causal-locality condition takes (CBCL - Safe Self-Extending Agent Communication, Endpoint Projection).

BSPL - The Blindingly Simple Protocol Language

Information-Driven Interaction-Oriented Programming: BSPL, the Blindingly Simple Protocol Language

Reference: Munindar P. Singh (2011). Proc. 10th International Conference on Autonomous Agents and Multiagent Systems (AAMAS 2011), Taipei, Taiwan, May 2–6, 2011, pp. 491–498. IFAAMAS. URL. PDF

Summary

Singh introduces BSPL, a declarative language for multiagent interaction protocols built from exactly two constructs: a message schema (a single message is an atomic protocol) and the composition of existing protocols. The paper’s programme is interaction-oriented programming (IOP): treat the protocol as an engineering abstraction in its own right, specify it in terms of the information it needs in order to proceed and the information it produces when enacted, and let agents be independently designed and operated, judged only by their participation in the specified interactions. BSPL states no constraints on the ordering or occurrence of messages. Every ordering and exclusion requirement that a traditional protocol notation would express with control-flow or choice operators is instead derived from the information model of the messages.

The mechanism is parameter adornment. Each protocol and each message carries parameters, and each parameter is adorned ⌜in⌝ (its binding must come from outside, that is from a prior message), ⌜out⌝ (its binding originates here, with declarative force) or ⌜nil⌝ (the sender must not know a binding at emission time). Some parameters are marked as the key: a key binding identifies an enactment, and any parameter may be bound at most once per key binding. From these follow BSPL’s three commitments to the relational view of protocols as tables of enactments: uniqueness (one enactment per key), integrity (every public parameter bound in a complete enactment) and immutability (bindings never change, which is what makes the design robust to asynchrony). Ordering emerges as ⌜out⌝-before-⌜in⌝ on a shared parameter; mutual exclusion emerges as two messages both adorning the same parameter ⌜out⌝ for the same key; conventional orderings that carry no information (pay before delivery) are captured by introducing a token parameter. Composition renames parameters through references, so the foreign-exchange case study (bilateral price discovery composed with itself into multilateral price discovery, from Desai et al.) needs no access to the constituents’ internals, unlike the AUML nesting and data-flow-axiom approaches it is compared against.

The semantics is sketched rather than formalised (that is the work of Semantics and Verification of Information-Based Protocols). A role’s history is the sequence of messages it has sent or received; a history vector collects one history per role; it is quiescent when every sent message has been received and viable when every emission was made by a role that knew all ⌜in⌝ bindings and no ⌜out⌝ or ⌜nil⌝ binding at the time. A protocol’s intension is its set of quiescent viable history vectors. Three conflict kinds organise correctness: ⌜out⌝–⌜in⌝ (ordering, handled by causality), ⌜nil⌝–⌜in⌝ and ⌜nil⌝–⌜out⌝ (knowledge, local to one role) and ⌜out⌝–⌜out⌝ (occurrence, which has no general solution but can be checked to be under one role’s control). Meaning is deliberately separated from structure: a means clause may attach a commitment to a message, and BSPL is offered as the operational substrate that commitment-based approaches such as Commitment Machines - Yolum and Singh previously lacked.

Key Ideas

Two constructs only. A protocol is one message schema or a composition of protocols; there is no distinction between atomic and composite protocols, and no sequence, choice or loop operator.
Information orientation and explicit causality. Every causal dependency is an information dependency: a message can be sent only when its ⌜in⌝ parameters are already bound in the sender’s local history. There are no hidden flows of causality because there are no hidden flows of information.
Adornments ⌜in⌝ / ⌜out⌝ / ⌜nil⌝ and keys. Adornments are relative to the protocol, not to a role. A top-level enactable protocol adorns all public parameters ⌜out⌝; a protocol with a public ⌜in⌝ parameter is not enactable on its own and must be composed.
Uniqueness, integrity, immutability. Enactments are immutable tuples keyed by the key parameters; a parameter bound twice for the same key is an integrity violation, which is how exclusive choice is expressed.
Shared-nothing enactment. No global state or repository is needed: everything relevant to the social state of an interaction is in the parameter bindings of the messages exchanged. Each agent acts from its local history alone.
Ordering and exclusion are derived. ⌜out⌝ precedes ⌜in⌝ on the same parameter; two ⌜out⌝ adornments of the same parameter make the messages mutually exclusive; ⌜nil⌝ forces a message before a role learns a binding.
Specification patterns. Duplicating a parameter, generating an identifier, local (hidden) parameters, standing offers with composite keys (Insurance Claims), flexible sourcing of ⌜out⌝ parameters, in-out polymorphism (an unadorned parameter accepts either direction), forwarding, mixed initiative, and digressions in the sense of Yolum and Singh.
Composition without breaking encapsulation. Multilateral price discovery is a composition of the generalised bilateral protocol with itself, with adornments in the two references doing the coordination work that Desai et al.’s data-flow axioms and Odell et al.’s AUML nesting required.
Comparison with procedural notations. Choreographies (WS-CDL, message sequence charts), AUML and RASA specify orderings directly; BSPL avoids the enactability failures Desai and Singh call blindness because the only way to state an ordering is a causally sound ⌜out⌝-to-⌜in⌝ flow. Nonlocal choice is not automatically avoided but is analysable.
Meaning stays separate. means clauses attach commitments to messages; BSPL supplies the operational underpinning that meaning-based protocol approaches assumed.
Named future work. Multiple keys, compositional verification of enactability, multicast (several agents playing one role) and discovery protocols with late role binding. Kohei Honda is thanked in the acknowledgements, a direct line to the Session Types tradition.

Connections

BSPL (concept hub)
Information Protocols
Enactability
Interaction Protocols
Semantics and Verification of Information-Based Protocols: the formal semantics and verifier for this language
Kiko - Programming Agents to Enact Interaction Protocols: the programming model built on it
Commitment Machines - Yolum and Singh
Flexible Protocol Specification and Execution
Commitment-Based Protocol
Commitment-based Semantics
ACL Rethinking Principles: the same author’s earlier argument for public, protocol-based ACL semantics
FIPA-ACL Specifications: the ordering-based interaction protocols BSPL positions itself against
Coordinating Agents Using ACL Conversations
Endpoint Projection
Session Types
Multiparty Asynchronous Session Types
Choreographic Programming
Knowledge of Choice
Realizability and Verification of MSC Graphs: the nonlocal-choice problem BSPL inherits from message sequence charts
Message Sequence Charts - A Survey
Declarative Specification
CALM Theorem: monotone local knowledge as the basis for coordination-free enactment
CBCL - Safe Self-Extending Agent Communication: the nearest contemporary relative; see the contrast on BSPL

Conceptual Contribution

Claim: Multiagent protocols can be specified without any control-flow constructs. Modelling messages by the information they require and produce, with adornments and keys, yields all necessary ordering and exclusion constraints, supports shared-nothing asynchronous enactment from local knowledge, composes without breaking encapsulation, and leaves meaning to a separate layer.
Mechanism: Message schemas with ⌜in⌝/⌜out⌝/⌜nil⌝ parameter adornments and key parameters; protocol composition by reference with parameter renaming; a relational reading of enactments (uniqueness, integrity, immutability); a sketched semantics of quiescent viable history vectors; a conflict taxonomy (ordering, knowledge, occurrence); a foreign-exchange case study against Desai et al.’s commitment-protocol formalisation.
Concepts introduced/used: Information Protocols, BSPL, Enactability, Interaction Protocols, Declarative Specification, Commitment-based Semantics, Knowledge of Choice, Realizability
Stance: language design / foundational
Relates to: Continues Singh’s programme from ACL Rethinking Principles (public, verifiable protocol semantics instead of mental states) and supplies the operational layer under Commitment Machines - Yolum and Singh and Flexible Protocol Specification and Execution. It is the multiagent-systems counterpart to Multiparty Asynchronous Session Types and Choreographic Programming: where those derive per-role behaviour from a global type by Endpoint Projection and presuppose a compiler and ordered channels, BSPL has no projection step at all, because every role acts directly on the global protocol using only its own history over an unordered transport. CBCL - Safe Self-Extending Agent Communication later arrives at a similar causal, coordination-free protocol layer from the LangSec side, with content-addressed predecessors and signatures in place of parameter keys.

Heterogeneity Resolution

The general problem, central to the ontology-change literature, of reconciling differences across ontologies at multiple levels: lexical, structural, semantic, and pragmatic. Subsumes alignment, matching, mapping, merging, and translation.

In this vault

Interaction Protocols

Specifications of legal conversation patterns above the message level. Three families: ordering-based (Petri nets, state machines, AUML, FIPA interaction protocols), type-based (Session Types, Choreographic Programming with Endpoint Projection), and information-based (BSPL Information Protocols, where ordering and exclusion are derived from what each message requires and produces). Commitment-Based Protocols layer meaning over any of them.

In this vault

Local State Transfer

(page does not exist)

BDI

Belief-Desire-Intention

Dominant agent-architecture paradigm: agents maintain beliefs, desires, and intentions and deliberate over them.

In this vault

BSPL

The Blindingly Simple Protocol Language, Munindar Singh’s declarative language for multiagent interaction protocols (AAMAS 2011) and the reference instance of Information Protocols. A BSPL protocol is a set of message schemas and references to sub-protocols; every parameter is adorned ⌜in⌝ (must already be known to the sender), ⌜out⌝ (bound, with declarative force, by this message) or ⌜nil⌝ (must not yet be known), and a subset of parameters forms the key that identifies an enactment. There are no sequence, choice or loop operators: ordering follows from ⌜out⌝-before-⌜in⌝ on a shared parameter, and exclusive choice from the rule that a parameter is bound at most once per key. Enactment is shared-nothing and asynchronous: each agent decides what it may send from its own history alone, over a transport that need be neither ordered nor reliable, with no central coordinator.

The programme runs from language (BSPL - The Blindingly Simple Protocol Language) through formal semantics and verification of Enactability, safety and liveness (Semantics and Verification of Information-Based Protocols) to programming models for agents (Kiko - Programming Agents to Enact Interaction Protocols, preceded by Stellar, Protocols over Things, Bungie and Mandrake in the same group). The LoST middleware (local state transfer, 2011) is the architectural style; Splee (2017) extends the language with sets and expressions; Tosca (2017) and Clouseau (2020) compile commitment specifications down to information protocols; Chopra, Christie and Singh’s JAIR 2020 evaluation of protocol languages is the group’s own comparative survey.

Relation to session types and choreographies. Multiparty Asynchronous Session Types and Choreographic Programming derive per-role behaviour from a global type by Endpoint Projection and presuppose a compiler, fixed roles and ordered channels. BSPL reaches decentralised enactment without a projection step: every role acts on the global protocol using only its local history, and enactability plays the part that projectability and Knowledge of Choice play on the session-type side. It shares the nonlocal-choice problem with message sequence charts (Realizability and Verification of MSC Graphs) but detects it as an empty intension.

Relation to CBCL. CBCL - Safe Self-Extending Agent Communication’s R5 causal protocols and its later role layer are the nearest contemporary relative: declarative, asynchronous, decentralised verification from local history with no ordering assumptions, in Singh’s public-commitments tradition (ACL Rethinking Principles). The differences are what CBCL adds: dependencies are content-hash edges to concrete message instances rather than parameter values, the threat model is adversarial with roles bound by nomination and signature, the message language is bounded to DCFL, dialects install at runtime, and the local-to-global correspondence is a mechanised theorem rather than a tool-checked property. Any comparison with session types owes BSPL the same paragraph.

In this vault

BSPL - The Blindingly Simple Protocol Language (Singh, AAMAS 2011)
Semantics and Verification of Information-Based Protocols (Singh, AAMAS 2012)
Kiko - Programming Agents to Enact Interaction Protocols (Christie, Singh & Chopra, AAMAS 2023)
Information Protocols
Enactability
Interaction Protocols
Commitment Machines - Yolum and Singh
Commitment-Based Protocol
Commitment-based Semantics
ACL Rethinking Principles
Endpoint Projection
Session Types
Choreographic Programming
Knowledge of Choice
CALM Theorem
CBCL - Safe Self-Extending Agent Communication

Information Protocols

Interaction protocols specified purely by the information that messages require and produce, with no control-flow constructs. Introduced by Singh with BSPL: each message schema names a sender role, a receiver role and parameters adorned ⌜in⌝, ⌜out⌝ or ⌜nil⌝, and a key subset of parameters identifies the enactment. Two constraints do all the work. Information causality: a message may be emitted only when its ⌜in⌝ parameters are already bound in the sender’s local history, so ordering is derived rather than stated. Information integrity: within one enactment (one key binding) a parameter may be bound at most once, so exclusive choice is derived rather than stated. Because both constraints are evaluated against local history and bindings are immutable, enactment is coordination-free and tolerates unordered, lossy delivery; a received message is either consistent with the local store and added, or rejected.

The paradigm’s correctness vocabulary is Enactability (some viable path completes the enactment), safety (no parameter bound twice per key across all roles) and liveness (completion always remains possible). Its programming counterpart is the form, an enabled message with its ⌜in⌝ parameters filled and its ⌜out⌝ parameters awaiting a decision (Kiko - Programming Agents to Enact Interaction Protocols). Meaning is layered separately: a means clause may attach a commitment to a message, connecting information protocols to Commitment-based Semantics.

Contrast with ordering-based specifications (FIPA-ACL interaction protocols, AUML, message sequence charts, WS-CDL) and with type-based ones (Session Types, Choreographic Programming), which fix sequences or derive local behaviour by Endpoint Projection. CBCL - Safe Self-Extending Agent Communication’s causal protocols are the content-addressed, adversarial-setting relative of this idea.

In this vault

End-to-End Arguments in System Design

Reference

Saltzer, J. H., Reed, D. P., & Clark, D. D. (1984). “End-to-End Arguments in System Design.” ACM Transactions on Computer Systems, 2(4), 277-288. URL

Summary

Saltzer, Reed, and Clark articulate a design principle for layered distributed systems that had long been used but rarely stated explicitly: functions requiring knowledge and action at the endpoints of a communication — such as reliable delivery, integrity checking, encryption, duplicate suppression — cannot be fully and correctly implemented at lower layers. Lower-layer implementations are at best performance optimizations; the end-to-end argument says they cannot substitute for the end-level check.

The canonical example is careful file transfer between two hosts. Even if the communication network offers reliable delivery, threats remain — disk errors at either host, memory corruption during buffering, software bugs in the file-transfer program itself. No amount of reliability layered into the network can defend against these; only an end-to-end checksum computed from the file on disk at host A and verified against the file on disk at host B closes the loop. The paper then iterates the argument through encryption (only the endpoints know the plaintext), duplicate suppression (only the application knows what “duplicate” means at the transaction level), delivery acknowledgements, and crash recovery.

The principle is a design heuristic, not an absolute rule: performance sometimes justifies redundant lower-layer mechanisms (e.g., per-hop error correction in a very noisy link). But it inverts the naïve “make the network as reliable as possible” instinct, provides the intellectual backbone for the Internet’s dumb-network / smart-edges architecture, and underwrites TCP’s placement in the hosts rather than the routers. Its influence extends to REST’s principled avoidance of server-side session state, to security architectures that refuse to trust intermediaries, and to the “fate-sharing” style of protocol design.

Key Ideas

End-to-end argument: a function that must be correct at endpoints cannot be completely implemented below the endpoints.
Lower layers as optimization: partial lower-level help is only a performance enhancement, never a correctness substitute.
Careful file transfer: the worked example — only an end-to-end checksum protects against all failure modes.
Dumb core, smart edges: Internet architecture as the principle’s canonical application.
Encryption placement: true confidentiality requires endpoint encryption; network-level encryption is not enough.
Acknowledgements: application-meaningful acks (e.g., “request served”) require endpoint involvement.
Cost-benefit nuance: redundancy below is justified when error rate or cost of retry makes it worthwhile.

Connections

Principled Design Of The Modern Web Architecture — Fielding’s REST thesis formalizes many end-to-end commitments.
REST
LangSec — input parsing at the application boundary is itself an end-to-end verification.
Actor Model — supervisor-style recovery relies on end-to-end state ownership.
Impossibility of Distributed Consensus with One Faulty Process — endpoints cannot delegate liveness to lower layers either.

Conceptual Contribution

Multi-Agent Systems

Systems of multiple autonomous agents that interact, coordinate, and sometimes compete.

Foundations

Intelligent Agents Theory and Practice — Wooldridge
Multiagent Systems Sycara
Agent-Oriented Programming — Shoham

Coordination & robustness

Summary

Shoham introduces Agent-Oriented Programming (AOP) as a specialization of object-oriented programming in which the state of each module — now called an agent — is a mental state composed of beliefs, capabilities, decisions/choices, commitments, and obligations. Interpretation of these components is formally grounded in an extension of standard epistemic logics that adds temporality, obligation, decision, and capability operators.

Agents communicate via speech-act-inspired primitives (inform, request, offer, promise, decline, etc.), constrained by rules such as honesty and consistency. The paper defines a class of agent interpreters and presents AGENT-0, a concrete implemented language whose semantics is tied to the mental-state logic. Shoham also discusses “agentification” — converting arbitrary devices into AOP-programmable agents — and situates AOP against BDI architectures, speech act theory, and McCarthy’s Elephant2000.

Key Ideas

AOP as OOP specialization with mental-state components.
Agents constrained by honesty/consistency on communicative acts.
Speech-act-typed primitives as message types.
AGENT-0 language and interpreter.
“Agentification” of arbitrary devices.

Connections

Conceptual Contribution

Claim: Programming can be fruitfully specialised to an “agent” paradigm in which module state is a formally specified mental state and inter-module communication is a speech act constrained by rational/ethical rules.
Mechanism: Shoham extends standard modal-epistemic logic with temporal, obligation, decision and capability operators, defines the mental-state components of an agent (belief, capability, commitment, choice, obligation), then specifies a generic agent interpreter cycle (update mental state, execute commitments, handle incoming messages) and instantiates it in AGENT-0, whose programs are condition-action rules over mental state and whose messages are typed INFORM/REQUEST/UNREQUEST primitives.
Concepts introduced/used: Agent-Oriented Programming, Mental State, Commitment, AGENT-0, Speech Act Theory, Agentification, BDI, Honesty Constraint
Stance: foundational
Relates to: Establishes the mentalistic ACL stance later criticised by Agent Communication Languages - Rethinking the Principles and surveyed by Intelligent Agents Theory and Practice; AGENT-0’s speech-act primitives anticipate KQML as an Agent Communication Language and FIPA-ACL; the mental-state architecture is reused in An Interaction-oriented Agent Framework for Open Environments.

Tags

Speech Acts

Austin and Searle’s theory that utterances are actions performed in speaking — assertions, requests, promises, declarations — each defined by a force plus propositional content and by felicity conditions. Speech-act theory is the linguistic foundation of both KQML and FIPA performatives.

In this vault

Commitment-based Semantics

Approach that grounds ACL message meaning in public social commitments rather than private mental states — making conformance verifiable.

In this vault

ACL Rethinking Principles
Agent Communication Languages - Rethinking the Principles
Agent Communication And Institutional Reality
A Common Ontology Of ACLs
Public Semantics
Mentalistic Semantics
Flexible Protocol Specification and Execution
A Commitment-Based Approach to Agent Communication
CBCL - Safe Self-Extending Agent Communication — modern LangSec-grounded ACL inheriting Singh’s social-agency programme; dialect installation as public commitment

FIPA-ACL

The Foundation for Intelligent Physical Agents Agent Communication Language — a standardised successor to KQML, defining performatives with mentalistic semantics (beliefs/intentions of sender and receiver).

Primary source

FIPA-ACL Specifications — the canonical standards documents (FIPA00037 Communicative Act Library, FIPA00061 Message Structure)
Communicative Actions for Artificial Agents - Cohen Levesque — Cohen & Levesque 1995, the direct semantic ancestor

Discussed in

Agent Communication Languages
Speech Act Theory
Multi-Agent Systems
Mentalistic Semantics
Public Semantics
Commitment-based Semantics
An Ontology for Commitments in Multiagent Systems - Singh
CBCL - Safe Self-Extending Agent Communication — DCFL-bounded ACL with public-commitment semantics, addressing the unverifiability of FIPA-ACL’s mentalistic SL semantics

KQML

Knowledge Query and Manipulation Language — one of the first ACLs, developed in the early 1990s by the ARPA Knowledge Sharing Effort. Performative-based, separating content language from communication.

Primary sources

Agent Communication Languages
FIPA-ACL — the standardised successor
Speech Act Theory
Ontolingua Portable Ontology Specifications — companion Knowledge Sharing Effort project
A Common Ontology Of ACLs
Verifiable Semantics for ACLs
CBCL - Safe Self-Extending Agent Communication — modern DCFL-bounded ACL with the same performative/extensibility ambitions but formal safety guarantees

An Interaction-oriented Agent Framework for Open Environments

An Interaction-oriented Agent Framework for Open Environments (Mercurio)

Reference: Baldoni, Baroglio, Bergenti, Marengo, Mascardi, Patti, Ricci, Santi (2011). AIIA 2011 (Italian AI Association)*. Source file: cbcl-ref/AI_IA2011.pdf. URL

Summary

The authors propose Mercurio, an agent-programming framework for open multi-agent systems that unifies direct (speech-act) and indirect (environment-mediated) communication under a single social/observational semantics based on commitments. Mercurio has three levels — specification (constitutive and regulative rules over a domain model), programming abstractions (Agents and Artifacts in the A&A meta-model), and infrastructure (built on JaCaMo integrating Jason, CArtAgO, and MOISE).

Commitment-based protocols give a public, verifiable meaning to actions (replacing mentalistic ACL semantics); constitutive rules define what counts as what, while regulative rules impose temporal constraints on the social state. Artifacts reify interaction patterns and the social state itself, enabling the environment to monitor and detect violations. The framework thus conjugates MAS flexibility with software-engineering modularity and compositionality, targeting e-government, cross-business, and service-oriented applications.

Key Ideas

Commitment-based social semantics replaces mentalistic ACL semantics.
Three levels: specification, programming abstractions, infrastructure.
Separation of constitutive and regulative rules.
Artifacts (A&A / CArtAgO) as first-class environment entities.
JaCaMo stack (Jason + CArtAgO + MOISE) as reference infrastructure.

Connections

Conceptual Contribution

Claim: Open multi-agent systems need a single, publicly verifiable interaction semantics that unifies direct (speech-act) and indirect (environment-mediated) communication; commitments plus Agents-and-Artifacts supply it and can be layered onto an existing AOP stack.
Mechanism: Mercurio stratifies into three levels. (1) Specification: a domain model plus constitutive rules (“what counts as what”) and regulative rules (temporal constraints on the social state). (2) Programming abstractions: the A&A meta-model (agents + artifacts) where artifacts reify interaction patterns and the social state itself. (3) Infrastructure: the JaCaMo stack (Jason + CArtAgO + MOISE) runs the above so violations are environment-detectable and the social state is inspectable.
Concepts introduced/used: Mercurio Framework, Commitment-Based Protocol, Constitutive Rules, Regulative Rules, Agents and Artifacts, JaCaMo, Jason, CArtAgO, MOISE, Social State
Stance: engineering
Relates to: Concretely instantiates the social-semantic programme of Agent Communication Languages - Rethinking the Principles and the open-issue agenda of Trends in Agent Communication Language and The State of the Art in Agent Communication Languages; its environment-reified interaction complements the self-organisation of A Composite Self-organisation Mechanism in an Agent Network.

Tags

CArtAgO

Common Artifacts for Agents Open framework — a Java infrastructure implementing the Agents & Artifacts meta-model in which agents perceive and act on first-class environmental artifacts governed by observable properties and operations.

In this vault

JaCaMo

An integrated multi-agent programming platform combining Jason (AgentSpeak BDI agents), CArtAgO (environments as artifacts), and Moise (organisational structures), instantiating the Agents & Artifacts metamodel. Cited as an interaction-oriented framework for open MAS.

In this vault

Modeling Rational Agents within a BDI-Architecture

Reference: Rao, A. S., & Georgeff, M. P. (1991). In J. Allen, R. Fikes, & E. Sandewall (Eds.), Proceedings of the Second International Conference on Principles of Knowledge Representation and Reasoning (KR’91), 473–484. Morgan Kaufmann. Also AAII Technical Note 14. Source file: rao_georgeff_bdi_architecture.pdf. URL

Summary

Rao and Georgeff give the canonical formalization of the Belief–Desire–Intention architecture as a modal/temporal logic over a branching-time possible-worlds model (a first-order extension of CTL*). Beliefs, goals (chosen, consistent, believed-achievable desires), and intentions are each modelled as accessibility relations over time trees, with the crucial innovation that belief-, goal-, and intention-accessible worlds are themselves time trees rather than static worlds — so an agent’s mental state simultaneously ranges over uncertainty about the world (branches across belief-worlds) and choice among options (branches within each world).

The paper realises the central tenets of Bratman’s philosophical theory (intentions as partial plans, irreducible to belief+desire) in a tractable semantic framework and introduces the distinction between choice (what the agent picks) and possibility (what the environment affords). A key technical contribution is the family of commitment strategies — blind, single-minded, and open-minded — which parametrically define how intentions persist under changing beliefs, yielding a taxonomy of rational agent types from fanatically committed to fully reconsidering. The formalism relates intentions to actions via sub-world inclusion (goals ⊆ beliefs; intentions ⊆ goals), avoiding classical side-effect problems of earlier intention logics.

Key Ideas

First-order CTL*-based modal logic with Bel, Goal, Intend accessibility.
Belief-, goal-, intention-accessible worlds are themselves branching time trees.
Optional vs. inevitable path modalities; distinguishes choice from possibility.
Strong realism constraint: goal-worlds are sub-worlds of belief-worlds.
Commitment strategies: blind, single-minded, open-minded.
Intentions treated as first-class, non-reducible to beliefs and desires.
Comparison with Cohen–Levesque’s reductive intention formalism.

Connections

Conceptual Contribution

Claim: Rational agent behaviour is best modelled by treating beliefs, goals and intentions as equally fundamental modal attitudes over branching time, with agent types differentiated by commitment strategy rather than by reduction of intention to other attitudes.
Mechanism: (i) Adopt CTL*-style branching time where each situation is a time point in a time tree, with path modalities optional and inevitable; (ii) associate each situation with sets of belief-, goal-, and intention-accessible worlds, each itself a time tree, imposing sub-world containment (strong realism); (iii) define rationality postulates relating the three attitudes; (iv) axiomatise commitment policies (blind = drop intention only on achievement, single-minded = also drop when no longer believed achievable, open-minded = also drop when no longer desired), yielding distinct agent classes.
Concepts introduced/used: BDI Logic, Commitment Strategies, Strong Realism, Branching Time, Intention, Practical Reasoning
Stance: formal framework
Relates to: Formalises the philosophical picture of Two Faces of Intention and operationalises the stance of Intentional Stance; provides the semantic bedrock for BDI Architecture, Agent-Oriented Programming and the theory side of Intelligent Agents Theory and Practice; differs from and complements Intention Is Choice with Commitment (Cohen & Levesque) by treating intention as primitive rather than derived.

Tags

BDI Architecture

The Belief-Desire-Intention model of practical reasoning: an agent maintains beliefs about the world, desires (goals) it would like to achieve, and intentions (committed plans) that drive its behaviour, typically executed via a sense-deliberate-act loop over a plan library.