LLM Agents

Large-language-model-powered agents: natural-language coordination, tool use, multi-agent orchestration.

Surveys & frameworks

Protocols & communication

Failures & threats

Lineage

Backlinks

Generative Agents line 18
ClawWorm Self-Propagating Attacks Across LLM Agent Ecosystems line 24
Augmented Language Models - A Survey line 27
Augmented Language Models - A Survey line 21
Survey Of AI Agent Protocols line 27
Survey Of AI Agent Protocols line 25
Survey Of AI Agent Protocols line 22
Multi-Agent Systems line 20
Toolformer line 28
Toolformer line 19
Toolformer line 8
Symbolic Plans line 7
Language Models are Few-Shot Learners line 28
Language Models are Few-Shot Learners line 18
Language Models are Few-Shot Learners line 6
Survey Of Agent Interoperability Protocols line 28
Survey Of Agent Interoperability Protocols line 23
Why Do Multi-Agent LLM Systems Fail line 27
Why Do Multi-Agent LLM Systems Fail line 18
Multi-Agent Collaboration in AI - Wasif Tunkel line 26
Multi-Agent Collaboration in AI - Wasif Tunkel line 18
Beyond Self-Talk - Communication-Centric Survey Of LLM Multi-Agent Systems line 30
Beyond Self-Talk - Communication-Centric Survey Of LLM Multi-Agent Systems line 21
Security Applications Of Formal Language Theory line 35
Intent Formalization - A Grand Challenge for Reliable Coding line 19
A Scalable Communication Protocol for Networks of LLMs line 30
A Scalable Communication Protocol for Networks of LLMs line 22
index line 19
Conversational AI line 7
Retrieval-Augmented Generation line 8
The Rise and Potential of LLM-Based Agents line 28
The Rise and Potential of LLM-Based Agents line 18
Language Agents line 7
Emergent Communication line 15
SoK The Attack Surface of Agentic AI line 24
Ripple Effect Protocol line 27
Ripple Effect Protocol line 21
Standard Operating Procedures (SOPs) line 7
Exploit Programming - From Buffer Overflows To Weird Machines line 31
Agent Network Protocol line 10
Agent Communication Languages line 35
Multi-Agent Collaboration Mechanisms - Survey of LLMs line 29
Multi-Agent Collaboration Mechanisms - Survey of LLMs line 20
AutoGen - Multi-Agent Conversation Framework line 29
AutoGen - Multi-Agent Conversation Framework line 19
MetaGPT Meta Programming for Multi-Agent Collaboration line 30
MetaGPT Meta Programming for Multi-Agent Collaboration line 18
On the Pitfalls of Measuring Emergent Communication line 19
Agent Hub line 7
Agent-to-Agent Protocol line 12
Edge Intelligence line 10
Edge Intelligence Survey line 31
Edge Intelligence Survey line 19
Distributed Security line 19
Chain-of-Thought Prompting line 27
Chain-of-Thought Prompting line 18
Chain-of-Thought Prompting line 8
Agents Framework - Zhou et al line 27
Agents Framework - Zhou et al line 18
CAMEL Communicative Agents for Mind Exploration of LLM Society line 29
CAMEL Communicative Agents for Mind Exploration of LLM Society line 18
From Eliza to XiaoIce - Social Chatbots line 18
Hallucination line 7
Why AI Agents Communicate In Human Language line 26
Why AI Agents Communicate In Human Language line 24
Why AI Agents Communicate In Human Language line 20
Levels Of Social Orchestration line 26
Levels Of Social Orchestration line 20
Trust and Reputation line 12
MalTool Malicious Tool Attacks line 26
MalTool Malicious Tool Attacks line 24
MalTool Malicious Tool Attacks line 20
Vibe Coding line 7
Large Population Models line 7
Inter-Agent Trust Models - A Comparative Study line 26
Long-short Term Memory line 7
AI Agents Under Threat line 26
AI Agents Under Threat line 24
AI Agents Under Threat line 18
Model Context Protocol line 12
Cicero Human-Level Play in Diplomacy line 31
Cicero Human-Level Play in Diplomacy line 18
Agents of Chaos line 20
MCP Landscape Security Threats And Future Research Directions line 31
MCP Landscape Security Threats And Future Research Directions line 22

Linked Pages

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

Canonical Languages

KQML — original knowledge-query language
FIPA-ACL — standardised successor

Surveys & Principles

The State of the Art in Agent Communication Languages
Trends in Agent Communication Language
ACL Rethinking Principles — Singh, mental vs. social semantics
Agent Communication Languages - Rethinking the Principles
A Common Ontology Of ACLs
Toward Automated Evolution of ACLs
Towards Automating the Evolution of Linguistic Competence

Semantics

Conversations & Protocols

Domain instances

Common Business Communication Language

From Eliza to XiaoIce - Social Chatbots

From Eliza to XiaoIce: Challenges and Opportunities with Social Chatbots

Reference: Heung-Yeung Shum, Xiaodong He, and Di Li (2018). arXiv:1801.01957v2 (Frontiers of IT & EE). Source file: 1801.01957v2.pdf. URL

Summary

Survey of conversational AI that traces the arc from Eliza (1966) and Parry through ALICE, DARPA Communicator, and modern IPAs (Siri, Cortana, Alexa) to Microsoft’s social chatbot XiaoIce. The authors distinguish task-completion from chitchat systems, and argue that the next frontier — social chatbots — must integrate IQ (knowledge/reasoning/skills) with EQ (empathy, social skills, personality) to build long-term emotional connections with users.

They introduce conversation-turns per session (CPS) as a success metric for social chatbots and describe XiaoIce’s architecture (chat manager, core chat with retrieval+generation, visual awareness, skills) along with design principles around ethical response filtering and persona consistency.

Key Ideas

IQ + EQ integration as design principle for social chatbots
CPS (conversation-turns per session) as success metric
Core-chat pipeline: user understanding -> response generation with topic guidance
Image commenting as social (not just descriptive) task
Taxonomy: chitchat vs task-completion vs IPA vs social chatbot

Connections

Conceptual Contribution

Claim: Social chatbots must be designed around both intellectual quotient (IQ) and emotional quotient (EQ); their success metric is not task completion or Turing-test pass-rate but Conversation-turns Per Session (CPS), reflecting sustained emotional engagement.
Mechanism: Survey-style design brief: trace Eliza → Parry → Alice → DARPA Communicator → Siri → XiaoIce; propose an architecture with chat manager, core-chat (user understanding + response generation + personality + ethical filter), visual awareness, and skills; integrate empathy, consistent persona, topic guidance, and emotion tracking; use XiaoIce (100M+ users) as illustrative case.
Concepts introduced/used: Social Chatbots, Conversation-turns Per Session (CPS), Emotional Quotient, Core Chat Architecture, Empathetic Response Generation, XiaoIce, Conversational AI, Eliza, Turing Test, Retrieval-based Dialogue
Stance: engineering / design-principles
Relates to: Stands at the opposite pole from task-oriented ACLs (KQML as an Agent Communication Language, FIPA-ACL) — here the communicative goal is emotional belonging rather than knowledge exchange. Contrasts with Agents Framework - Zhou et al (task agents with SOPs) and Multi-Agent Collaboration in AI - Wasif Tunkel (productive collaboration). Anticipates the human-emotional concerns that re-emerge in Ensuring Trustworthy and Ethical Behaviour in Intelligent Logical Agents.

Tags

MalTool Malicious Tool Attacks

MalTool: Malicious Tool Attacks on LLM Agents

Reference: Hu, Jia, Li, Song, Gong (2026). arXiv:2602.12194 (Duke, UC Berkeley). Source file: 2602.12194v2.pdf. URL

Summary

This paper presents the first systematic study of code-level malicious tool attacks on LLM agent ecosystems (MCP, Skills, mcp.so, skillsmp). Whereas prior work focused on crafting misleading tool names and descriptions, the authors show that genuinely harmful behaviour must be embedded in a tool’s implementation. They propose a CIA (confidentiality/integrity/availability) taxonomy of 12 concrete malicious behaviours (data exfiltration, credential abuse, data poisoning, file deletion, RCE downloading, CPU/GPU hijacking, DoS).

They build MalTool, a coding-LLM framework that iteratively synthesizes standalone and Trojan malicious tools using a behaviour-specific system prompt, diversity guidance, and an execution-based verifier. The result: 1,200 standalone malicious tools and 5,287 real-world tools with injected malicious behaviours. Detection methods (VirusTotal, Cisco MCP Scanner, MCPScan) perform poorly, motivating new defences.

Key Ideas

CIA taxonomy of malicious tool behaviours in agent settings.
Automatic generation pipeline: system prompt + coding LLM + execution-based verifier.
Trojan construction by embedding malicious logic in benign tool code.
Existing malware and MCP-specific scanners fail on both false-positives and false-negatives.
Dataset released for benign tools only to minimize misuse.

Connections

Conceptual Contribution

Claim: Truly harmful behaviour in LLM Agents ecosystems lives in tool implementations, not in their descriptions; prior description-level red-teaming misses the dominant attack class, and current scanners miss it too.
Mechanism: Introduces a CIA taxonomy of 12 malicious behaviours; builds MalTool, a coding-LLM pipeline (behaviour-specific system prompt + diversity guidance + execution-based verifier) that produces standalone and Trojan tools; benchmarks VirusTotal, Cisco MCP Scanner, MCPScan and shows poor detection.
Concepts introduced/used: Tool Use, Model Context Protocol, Trojan Tools, Prompt Injection, Agent Security, LLM Agents, Distributed Security
Stance: empirical
Relates to: Deepens the tool/MCP threat surface catalogued in AI Agents Under Threat and Survey Of Agent Interoperability Protocols; motivates language-based defences akin to A Language-Based Approach To Prevent DDoS and capability isolation of Security Kernel Lambda Calculus.

Tags

AI Agents Under Threat

AI Agents Under Threat: A Survey of Key Security Challenges and Future Pathways

Reference: Deng, Guo, Han, Ma, Xiong, Wen, Xiang (2025). ACM Computing Surveys 57(7), Article 182. Source file: 3716628.pdf. URL

Summary

This survey organizes the emerging threat landscape of LLM-powered AI agents around four knowledge gaps: unpredictability of multi-step user inputs, complexity of internal execution, variability of operational environments, and interactions with untrusted external entities. It unifies single-agent and multi-agent attack surfaces within a perception/brain/action + agent2agent/agent2env/agent2memory taxonomy.

Concrete threats reviewed include adversarial prompts, prompt injection, jailbreaks, backdoor attacks, hallucination and misalignment, tool-use risks, indirect prompt injection, reinforcement-learning environment attacks, cooperative and competitive inter-agent risks, and long/short-term memory attacks. The authors tabulate defenses (prevention- and detection-based), rate their efficacy, and highlight open directions for robust and trustworthy agents.

Key Ideas

Four knowledge gaps framing agent security.
Taxonomy: perception / brain / action / agent2agent / agent2env / agent2memory threats.
Six categories of prompt-injection attack engineering (naive, escape, context-ignore, fake-completion, multimodal, combined).
Jailbreak domino effect in multi-agent populations.
Memory poisoning and indirect prompt injection as underexplored surfaces.

Connections

Conceptual Contribution

Claim: LLM Agents security should be organised around four knowledge gaps (input unpredictability, internal complexity, environmental variability, untrusted interactions) mapped onto a perception/brain/action + agent2{agent,env,memory} taxonomy.
Mechanism: Surveys adversarial prompts, prompt injection, jailbreaks, backdoors, hallucination, tool-use risks, indirect injection, RL environment attacks, inter-agent cooperative/competitive risks, memory poisoning; tabulates prevention- vs detection-based defences and rates their efficacy.
Concepts introduced/used: Prompt Injection, Jailbreak, Backdoor Attacks, Tool Use, Memory Poisoning, Hallucination, Model Context Protocol, LLM Agents, Multi-Agent Systems, Trust and Reputation, Distributed Security, Agent Security
Stance: survey
Relates to: Provides the threat scaffolding that MalTool Malicious Tool Attacks deepens at the tool layer; complements lifecycle threats in Survey Of Agent Interoperability Protocols; motivates static-analysis defences like A Language-Based Approach To Prevent DDoS.

Tags

Why Do Multi-Agent LLM Systems Fail

Why Do Multi-Agent LLM Systems Fail?

Reference: Mert Cemri, Melissa Z. Pan, Shuyi Yang, Lakshya A. Agrawal, Bhavya Chopra, Rishabh Tiwari, Kurt Keutzer, Aditya Parameswaran, Dan Klein, Kannan Ramchandran, Matei Zaharia, Joseph E. Gonzalez, Ion Stoica (2025). arXiv:2503.13657v2 (UC Berkeley). Source file: 2503.13657v2.pdf. URL

Summary

First empirically grounded taxonomy of failure modes in Multi-Agent LLM Systems (MAS). The authors analyse 200+ execution traces from seven popular MAS frameworks (MetaGPT, ChatDev, HyperAgent, AppWorld, AG2, Magentic-One, OpenManus), annotated by six human experts via grounded theory and reaching Cohen’s κ ≈ 0.88, and distil 14 fine-grained failure modes grouped into three categories: Specification Issues (42%), Inter-Agent Misalignment (37%), and Task Verification (21%).

They release MAST (Multi-Agent System failure Taxonomy), a validated LLM-as-judge pipeline for automated failure diagnosis, and two intervention case studies showing that architectural/prompt fixes inspired by MAST improve success rates modestly — demonstrating that MAS failures are system-design problems, not merely model-capability problems.

Key Ideas

Three failure categories: specification, inter-agent misalignment, verification
14 fine-grained failure modes including step repetition, information withholding, task derailment
Grounded-theory methodology with rigorous inter-annotator agreement (κ=0.88)
LLM-as-judge pipeline (MAST) achieves κ=0.77 vs humans for scalable evaluation
Insight: better specifications and verification beat bigger models

Connections

Conceptual Contribution

Claim: Multi-Agent LLM System (MAS) failures are predominantly system-design problems — specification, coordination, and verification — rather than base-model capability problems; and these failures have an empirically discoverable, reproducible taxonomy.
Mechanism: Grounded-theory analysis of 200+ execution traces across seven MAS frameworks (MetaGPT, ChatDev, HyperAgent, AppWorld, AG2, Magentic-One, OpenManus) with six human expert annotators; iterative refinement to Cohen’s κ≈0.88; yield the 14-mode MAST taxonomy grouped into Specification Issues (42%), Inter-Agent Misalignment (37%), Task Verification (21%); validate an LLM-as-judge annotator (κ≈0.77); intervention case studies showing prompt/architecture fixes provide only modest gains, motivating deeper redesign.
Concepts introduced/used: MAST Taxonomy, Grounded Theory, Inter-Agent Misalignment, LLM-as-judge, Specification Issues, Task Verification, Multi-Agent Systems, LLM Agents, Cohen’s Kappa, Standard Operating Procedures (SOPs)
Stance: empirical / evaluative
Relates to: Supplies empirical grounding for the design-quality concerns in Agents Framework - Zhou et al (SOPs attempt to mitigate FC1 specification issues) and Multi-Agent Collaboration in AI - Wasif Tunkel. Inter-agent misalignment mirrors the formal pathologies in Are Multiagent Systems Resilient to Communication Failures. Motivates richer communication protocols like A Scalable Communication Protocol for Networks of LLMs and commitment-style ACLs (ACL Rethinking Principles).

Tags

Levels Of Social Orchestration

Levels of Social Orchestration for Agentic Systems

Reference: Chopra, Bhattacharya, Leibo, Raskar (2025). ICML 2025 (MIT, Google DeepMind). Source file: agentic_draft.pdf. URL

Summary

The authors argue that as AI agents scale to billions, beneficial collective behaviour depends less on maximizing individual intelligence and more on discovering interaction protocols. They introduce Large Population Models (LPMs) - differentiable, end-to-end trainable protocols spanning simulated and physical agent networks - as a paradigm shift from LLMs (data -> language) to LPMs (protocol -> population).

They propose a five-level taxonomy of agentic systems: L1 Perceive, L2 Automate, L3 Connect (all within human cognitive bounds), then L4 Navigate and L5 Transform (beyond Dunbar-scale human coordination). Case studies span pandemic response, traffic coordination, and Coachella-style crowd management, framing the progression from information intelligence to collective orchestration.

Key Ideas

Protocol-centric intelligence: rules of interaction beat bigger individual models.
Large Population Models (LPMs): differentiable protocols over synthetic+physical agents.
L1-L5 levels: Perceive, Automate, Connect, Navigate, Transform.
Human Connectivity Barrier (~1500 people) as natural scaling limit.
Case studies in pandemics, traffic, crowd scheduling.

Connections

Conceptual Contribution

Claim: At population scale, beneficial AI comes from protocol design, not model scaling; the paradigm must shift from LLMs (data to language) to Large Population Models (protocol to population).
Mechanism: Introduces differentiable, end-to-end trainable LPMs spanning simulated and physical agents; proposes a 5-level taxonomy (Perceive/Automate/Connect within human cognitive bounds, Navigate/Transform beyond them); case studies in pandemic response, traffic, crowd management.
Concepts introduced/used: Large Population Models, Differentiable Protocols, Human Connectivity Barrier, Multi-Agent Systems, LLM Agents, Self-Adaptive Systems, Emergent Communication
Stance: foundational / engineering
Relates to: Provides the level-taxonomy lens under which Ripple Effect Protocol sits as an L4 mechanism; echoes the “protocols-not-messages” critique of Why AI Agents Communicate In Human Language; its macro vision contrasts with the infrastructural surveys Survey Of AI Agent Protocols and Survey Of Agent Interoperability Protocols.

Tags

Ripple Effect Protocol

Ripple Effect Protocol: Coordinating Agent Populations

Reference: Chopra, Sharma, Ahmad, Muscariello, Pandey, Raskar (2025). arXiv:2510.16572 (Project Iceberg, MIT / Cisco). Source file: 2510.16572v1.pdf. URL

Summary

REP is a coordination protocol for populations of LLM agents that augments existing messaging (A2A, ACP, SLIM) with sensitivity sharing: agents broadcast not only their decisions but lightweight signals expressing how those decisions would change under counterfactual environmental shifts. Neighbours aggregate these sensitivities into shared coordination variables, letting groups converge faster and more stably than with decision-only exchange.

The protocol separates cognition (local LLM policy) from coordination (aggregation + optional consensus). Experiments on the Beer Game (bullwhip reduction of 41.8%), Fishbanks (sustainability +25%), and movie-scheduling show 41-100% coordination-accuracy gains over A2A baselines and scale from 10 to 200 agents.

Key Ideas

Sensitivity = textual or numeric derivative of a decision w.r.t. environment.
Four-step round: receive, decide+sensitivity, aggregate neighbors, optional median consensus.
Modality-agnostic aggregator phi (numeric gradient or LLM-synthesized textual update).
Transport-agnostic: works over SLIM, A2A, ACP.
Mitigates information cascades / bullwhip effects in open agent networks.

Connections

Conceptual Contribution

Claim: Agent populations coordinate faster and more stably when they share not just decisions but sensitivities - counterfactual derivatives of decisions w.r.t. the environment.
Mechanism: A transport-agnostic four-step round (receive, decide+sensitivity, aggregate, optional median consensus) layered atop Agent-to-Agent Protocol/ACP/SLIM, validated on Beer Game (bullwhip -41.8%), Fishbanks, and movie scheduling up to 200 agents.
Concepts introduced/used: Sensitivity Sharing, Coordination Variables, Information Cascades, Gossip Protocols, Multi-Agent Systems, LLM Agents, Agent-to-Agent Protocol, Emergent Communication
Stance: engineering
Relates to: Realises the structured-coordination vision of Why AI Agents Communicate In Human Language and instantiates an L4 Navigate-level protocol from Levels Of Social Orchestration; extends aggregation intuitions from Gossip Protocols.

Tags

Why AI Agents Communicate In Human Language

Why do AI agents communicate in human language?

Reference: Zhou, Feng, Julaiti, Yang (2025). arXiv:2506.02739. Source file: 2506.02739v1.pdf. URL

Summary

The authors argue that natural language, inherited from single-agent LLM pretraining, is fundamentally misaligned with the needs of multi-agent coordination. Because LLMs are trained to maximize likelihood over discrete token sequences, their internal representations are high-dimensional and continuous, but their outputs are forced into a sparse, ambiguous, non-differentiable symbolic form that loses information when used as an inter-agent channel.

They formalize this as a semantic misalignment problem: cascading encode/decode cycles across agents accumulate lossy projection errors and prevent gradient flow. The paper calls for a new multi-agent modeling paradigm where agents coordinate via structured, learnable representations shaped by role persistence, state tracking, and explicit coordination graphs, rather than free-form natural-language dialogue.

Key Ideas

Natural language is a lossy, non-differentiable projection of LLM hidden states.
Cascading communication rounds accumulate semantic error.
Protocol-induced misbehavior: naive-literal interpretation and action-state decoupling.
Advocates structured message schemas, role-consistent embeddings, coordination graphs.
Critique of AutoGen, MetaGPT, CAMEL-style NL-based multi-agent frameworks.

Connections

Conceptual Contribution

Claim: Natural language is an accidental, lossy, non-differentiable channel for inter-LLM Agents coordination; multi-agent AI needs a purpose-built representational substrate.
Mechanism: Formalises repeated encode/decode cycles as error-accumulating projections from continuous hidden states to sparse tokens; diagnoses “protocol-induced misbehavior” (naive-literal reading, action-state decoupling); prescribes structured schemas, role-consistent embeddings, and explicit coordination graphs.
Concepts introduced/used: Semantic Misalignment, Emergent Communication, Coordination Graphs, Multi-Agent Systems, LLM Agents, Agent Communication Languages, Differentiable Protocols
Stance: critique
Relates to: Provides the theoretical motivation that Ripple Effect Protocol and Levels Of Social Orchestration operationalise; contrasts with the symbolic, performative-centric designs of KQML and FIPA-ACL by rejecting symbolic channels entirely.

Tags

A Scalable Communication Protocol for Networks of LLMs

A Scalable Communication Protocol for Networks of Large Language Models

Reference: Samuele Marro, Emanuele La Malfa, Jesse Wright, Guohao Li, Nigel Shadbolt, Michael Wooldridge, Philip Torr (2024). arXiv:2410.11905v1 (Oxford / Eigent AI). Source file: 2410.11905v1.pdf. URL

Summary

Introduces Agora, a meta-protocol for inter-agent communication in large heterogeneous networks of LLM-powered agents. Agora frames the design space as the Agent Communication Trilemma — versatility, efficiency, portability — and argues no single format (natural language, structured APIs like REST, or semantic-web RDF) can satisfy all three simultaneously.

Agora’s trick is to use different formats for different traffic volumes: rare/novel messages flow as natural language handled by LLMs; frequent patterns are formalised into Protocol Documents (PDs) negotiated between agents and then served by cheap LLM-written routines. A 100-agent demo shows emergent self-organising protocols and ~5x cost reduction over natural-language-only communication.

Key Ideas

Agent Communication Trilemma: versatility vs efficiency vs portability
Protocol Documents (PDs): hash-identified, agent-negotiated, machine-readable specs
Hybrid hierarchy: NL bootstrap -> PD negotiation -> LLM-written routines -> traditional protocols
Fully decentralised, hash-addressed storage (IPFS-compatible)
Emergent protocols among 100 heterogeneous LLM agents without central coordination

Connections

Conceptual Contribution

Claim: No single communication format can simultaneously satisfy versatility, efficiency, and portability (the Agent Communication Trilemma) at scale; a meta-protocol that dynamically mixes natural language, structured data, and LLM-written routines can sidestep the trilemma.
Mechanism: Agora uses hash-identified Protocol Documents (PDs) — plain-text, implementation-agnostic specs — negotiated on demand between LLM agents. Frequent traffic is handled by cheap LLM-written routines implementing a PD; rare or novel traffic falls back to LLMs with natural language. Decentralised, content-addressed (IPFS-style) PD distribution; demonstrated on a 100-agent heterogeneous network showing emergent protocols and ~5× cost reduction.
Concepts introduced/used: Agent Communication Trilemma, Protocol Documents, Meta-protocol, Emergent Protocols, Emergent Communication, LLM Agents, Content-addressed Storage, Negotiated Protocols, Negotiation, Agent Communication Languages
Stance: engineering / systems
Relates to: Modern successor to KQML as an Agent Communication Language and FIPA-ACL, replacing stipulated performatives with negotiated PDs. Echoes the emergent-language findings of Multi-Agent Cooperation and the Emergence of Natural Language and Emergence of Grounded Compositional Language in Multi-Agent Populations at the protocol-document level. Overlaps the design space of Model Context Protocol, Agent-to-Agent Protocol, Agent Network Protocol. Its bottom-up spirit mirrors The Extensible Language - Graham.

Tags

Multi-Agent Collaboration in AI - Wasif Tunkel

Multi-Agent Collaboration in AI: Enhancing Software Development with Autonomous LLMs

Reference: Mubeen Wasif and David Tunkel (2025). ResearchGate preprint. Source file: 16.pdf

Summary

A short survey-style paper arguing that multi-agent LLM systems can improve software-development workflows by distributing tasks (requirements, code generation, testing, documentation) across specialised autonomous agents that communicate via structured dialogue. The authors report qualitative experimental findings that division of labour and iterative refinement among agents produce higher-quality outputs than single-agent baselines.

The paper also surveys open challenges: coordination overhead, response consistency, bias propagation, and governance/security concerns. It advocates human-in-the-loop validation and explainability (XAI) as mitigations, and points to future integration with IDEs, CI/CD, and RAG.

Key Ideas

Specialised agent roles (coder, tester, documenter) mirror human dev teams
Structured inter-agent dialogue enables iterative code refinement
Hybrid human-AI teams recommended for reliability
Coordination cost, bias propagation, accountability are unsolved
RAG and adaptive prompting as future contextual-awareness tools

Connections

Conceptual Contribution

Claim: Multi-agent LLM systems, with role-specialised agents (coder, tester, documenter) communicating through structured dialogue, outperform single-agent LLMs on software-engineering tasks by mirroring human team division-of-labour.
Mechanism: Survey-plus-experiment discussion: assign distinct agents to requirement analysis, code generation, debugging, documentation; use iterative feedback loops and structured prompts; report efficiency/accuracy gains while flagging coordination overhead, bias propagation, and security concerns; advocate human-in-the-loop governance.
Concepts introduced/used: LLM Agents, Role-specialised Agents, Human-in-the-loop, Agent Coordination Overhead, Explainable AI, Multi-Agent Systems, Retrieval-Augmented Generation, Division of Labour
Stance: engineering / survey
Relates to: A light-weight practitioner view of the same problem space that Why Do Multi-Agent LLM Systems Fail tackles rigorously, and that Agents Framework - Zhou et al operationalises as an open-source library. Shares the communication-protocol concern with A Scalable Communication Protocol for Networks of LLMs.

Tags

Survey Of Agent Interoperability Protocols

A Survey of Agent Interoperability Protocols: MCP, ACP, A2A, and ANP

Reference: Ehtesham, Singh, Gupta, Kumar (2025). arXiv:2505.02279. Source file: 2505.02279v1.pdf. URL

Summary

This survey examines four emerging agent communication protocols targeting different interoperability tiers: the Model Context Protocol (MCP) for JSON-RPC tool invocation and context delivery; the Agent Communication Protocol (ACP) for REST-native multi-part performative messaging; the Agent-to-Agent Protocol (A2A) for peer-to-peer Agent-Card-based task outsourcing; and the Agent Network Protocol (ANP) for decentralized discovery using DIDs and JSON-LD.

The authors contrast architectures, discovery mechanisms, security models, and communication patterns, then recommend a phased adoption roadmap (MCP for tool access, then ACP for messaging, A2A for collaborative execution, ANP for open marketplaces). A timeline traces ancestry from KQML (1993) and FIPA-ACL (2000) through RAG, ReAct, function-calling up to modern agent protocols.

Key Ideas

Phased adoption roadmap: MCP -> ACP -> A2A -> ANP.
MCP core primitives: Tools, Resources, Prompts, Sampling under JSON-RPC 2.0.
A2A introduces Agent Cards, Tasks, Artifacts for enterprise-scale delegation.
ANP uses DIDs and JSON-LD for decentralized, internet-scale agent discovery.
Security threats tabulated across creation/operation/update lifecycle phases.

Connections

Conceptual Contribution

Claim: Modern agent interoperability is best understood as a four-tier stack (MCP for tools, ACP for messaging, A2A for delegation, ANP for open discovery) and should be adopted in that phased order.
Mechanism: Structured comparison of architectures, discovery, security, and message patterns; historical timeline rooting each protocol in KQML/FIPA-ACL ancestry; lifecycle threat table.
Concepts introduced/used: Model Context Protocol, Agent-to-Agent Protocol, Agent Network Protocol, Agent Communication Protocol, KQML, FIPA-ACL, Agent Cards, Decentralized Identifiers, JSON-RPC, Tool Use, LLM Agents
Stance: survey
Relates to: Complements the broader Survey Of AI Agent Protocols with a narrower, adoption-oriented roadmap. Its security-threat lifecycle connects directly to AI Agents Under Threat and MalTool Malicious Tool Attacks.

Tags

Survey Of AI Agent Protocols

A Survey of AI Agent Protocols

Reference: Yang, Chai, Song, Qi, Wen, Li, Liao, Hu, Lin, Chang, Liu, Wen, Yu, Zhang (2025). arXiv:2504.16736. Source file: 2504.16736v2.pdf. URL

Summary

This survey offers the first comprehensive classification and analysis of emerging AI agent protocols for LLM-based agents. The authors propose a two-dimensional taxonomy: (object orientation) context-oriented vs inter-agent protocols, and (application scenario) general-purpose vs domain-specific, covering MCP, A2A, ANP, ACP, Agora, LMOS, agents.json, LOKA, PXP, CrowdES, and others.

The paper then evaluates these protocols across efficiency, scalability, security, reliability, extensibility, operability, and interoperability, and sketches a forward-looking agenda: protocols should evolve from static to adaptive, from rules to ecosystems, and from mere communication to collective intelligence infrastructure.

Key Ideas

Two-dimensional taxonomy of agent protocols (object orientation x application scenario).
MCP as a universal context-oriented protocol with Host/Client/Server/Resource roles.
Inter-agent layer splits into general-purpose (A2A, ANP, AITP, ACP, Agora) and domain-specific (robot, human-computer, system).
Evaluation across 7 axes; case studies of MCP, A2A, ANP, Agora.
Next-generation protocols need adaptability, privacy preservation, group interaction.

Connections

Conceptual Contribution

Claim: The zoo of emerging LLM Agents protocols can be organised along two orthogonal axes (context-oriented vs inter-agent; general-purpose vs domain-specific), and evaluated on a shared seven-axis rubric.
Mechanism: Builds a taxonomy, then systematically compares Model Context Protocol, Agent-to-Agent Protocol, Agent Network Protocol, ACP, Agora, LMOS, agents.json, LOKA, PXP, CrowdES against efficiency, scalability, security, reliability, extensibility, operability, interoperability, with case studies.
Concepts introduced/used: Model Context Protocol, Agent-to-Agent Protocol, Agent Network Protocol, Agent Communication Languages, LLM Agents, Multi-Agent Systems, Interoperability, Agent Discovery
Stance: survey
Relates to: Shares its subject with Survey Of Agent Interoperability Protocols but takes a broader taxonomic view; its forward-looking “protocols as ecosystems” framing converges with Levels Of Social Orchestration and motivates coordination layers like Ripple Effect Protocol. Historical continuity with KQML and FIPA-ACL is implicit.

Tags

Agents Framework - Zhou et al

Agents: An Open-source Framework for Autonomous Language Agents

Reference: Wangchunshu Zhou et al. (2023). arXiv:2309.07870v3 (AIWaves, Zhejiang University, ETH Zürich). Source file: 2309.07870v3.pdf. URL

Summary

Introduces AGENTS, an open-source framework for building LLM-powered autonomous agents with first-class support for planning, long/short-term memory, tool use and web navigation, multi-agent communication, human-agent interaction, and fine-grained symbolic control via Standard Operating Procedures (SOPs). SOPs are state-graphs with LLM-editable transition rules and per-state prompt/tool configurations, giving users predictable, tunable control over otherwise stochastic agent behaviour.

The framework is declarative (agents instantiated from config JSON), supports dynamic scheduling of which agent speaks next in multi-agent settings, provides a FastAPI deployment target and an Agent Hub for sharing/forking agents, and includes an automated SOP-generation pipeline (meta-agent).

Key Ideas

SOP as a symbolic plan / state graph for controllable agents
Dynamic scheduling: LLM controller picks next actor rather than fixed order
Memory split: long-term (VectorDB + sentence-transformers) vs short-term scratchpad
Config-driven agent construction reduces boilerplate
Meta-agent auto-generates SOPs from task descriptions via RAG

Connections

Conceptual Contribution

Claim: Autonomous LLM-powered language agents become reliably controllable and customisable when their behaviour is specified by symbolic plans — Standard Operating Procedures (SOPs) — represented as state graphs that an LLM-based controller traverses, rather than by monolithic prompts alone.
Mechanism: Open-source library AGENTS unifying planning, long/short-term memory (VectorDB + scratchpad), tool use & web navigation, multi-agent communication (with LLM-moderator for dynamic scheduling), human-agent interaction (is_human flag), and controllability via SOPs. Includes an automated “meta-agent” that generates SOPs and configs from a task description via RAG.
Concepts introduced/used: Language Agents, Standard Operating Procedures (SOPs), Symbolic Plans, LLM Agents, Agent Hub, Dynamic Scheduling, Long-short Term Memory, Meta-agent, Retrieval-Augmented Generation, Tool Use, Human-in-the-loop
Stance: engineering / framework
Relates to: A concrete instantiation of the role-specialised multi-agent style advocated in Multi-Agent Collaboration in AI - Wasif Tunkel. Its SOP/controller discipline directly targets the failure modes later catalogued in Why Do Multi-Agent LLM Systems Fail. Its inter-agent messaging is more prescriptive than the negotiated protocols of A Scalable Communication Protocol for Networks of LLMs, sitting between classic ACLs (KQML as an Agent Communication Language, FIPA-ACL) and fully emergent LLM communication.