# Awesome Multi Agent Systems

> Curated, annotated list of multi-agent systems resources — MARL, game theory, negotiation, LLM agent teams, MCP/A2A protocols, benchmarks, and frameworks

- **Type:** MCP server
- **Install:** `agentstack add mcp-bloo-mind-awesome-multi-agent-systems`
- **Verified:** Yes — security-reviewed for prompt injection and unsafe behavior
- **Seller:** [bloo-mind](https://agentstack.voostack.com/s/bloo-mind)
- **Installs:** 0
- **Category:** [AI & ML](https://agentstack.voostack.com/c/ai-and-ml)
- **Latest version:** 0.1.0
- **License:** MIT
- **Upstream author:** [bloo-mind](https://github.com/bloo-mind)
- **Source:** https://github.com/bloo-mind/awesome-multi-agent-systems

## Install

```sh
agentstack add mcp-bloo-mind-awesome-multi-agent-systems
```

Requires the [AgentStack CLI](https://agentstack.voostack.com/docs/cli). Works with Claude Code, Cursor, and any MCP-compatible agent.

## About

# Awesome Multi-Agent Systems [](https://awesome.re)


[](LICENSE)
[](https://github.com/bloo-mind/awesome-multi-agent-systems/actions/workflows/link-check.yml)


Autonomous agents that cooperate, compete, negotiate, and learn — a curated, **annotated** guide to Multi-Agent Systems (MAS), from classical coordination theory and multi-agent reinforcement learning (MARL) to LLM-based agent teams. Every entry says what it is, why it matters, and when to use it.

**Start here:** [The essential dozen](#start-here-the-essential-dozen) · [Which framework should I use?](#which-framework-should-i-use) · [Find a benchmark](#datasets-and-benchmarks) · [Understand failure modes](#recent-high-impact-papers-and-surveys) · [Wire up agent protocols](#agent-interoperability-protocols)

**Something missing?** [Suggest a resource in two minutes](https://github.com/bloo-mind/awesome-multi-agent-systems/issues/new/choose) — or grab a [help-wanted task](https://github.com/bloo-mind/awesome-multi-agent-systems/issues?q=is%3Aissue+is%3Aopen+label%3A%22good+first+issue%22).

## Contents

- [Awesome Multi-Agent Systems ](#awesome-multi-agent-systems-)
  - [Contents](#contents)
  - [Start here: the essential dozen](#start-here-the-essential-dozen)
  - [Getting started: three pathways](#getting-started-three-pathways)
  - [Explore by topic](#explore-by-topic)
  - [Curated resources](#curated-resources)
    - [Classic books (pre-2024)](#classic-books-pre-2024)
    - [Recent books (2024–)](#recent-books-2024)
    - [Tutorials and courses](#tutorials-and-courses)
    - [How-to guides and framework docs](#how-to-guides-and-framework-docs)
    - [Seminal papers and milestone work](#seminal-papers-and-milestone-work)
    - [Recent high-impact papers and surveys](#recent-high-impact-papers-and-surveys)
    - [Datasets and benchmarks](#datasets-and-benchmarks)
    - [Frameworks, libraries, and tools](#frameworks-libraries-and-tools)
      - [Which framework should I use?](#which-framework-should-i-use)
      - [LLM-based multi-agent frameworks](#llm-based-multi-agent-frameworks)
      - [MARL, simulation, and classic MAS platforms](#marl-simulation-and-classic-mas-platforms)
    - [Competitions and challenges](#competitions-and-challenges)
    - [Agent interoperability protocols](#agent-interoperability-protocols)
  - [Reproducibility and community](#reproducibility-and-community)
    - [Reproducibility resources and code](#reproducibility-resources-and-code)
    - [Community resources](#community-resources)
  - [Field guide: taxonomy and milestone timeline](#field-guide-taxonomy-and-milestone-timeline)
    - [MAS topic taxonomy](#mas-topic-taxonomy)
    - [Milestone timeline](#milestone-timeline)
  - [Recently added](#recently-added)
  - [Editorial policy](#editorial-policy)
  - [Related awesome lists](#related-awesome-lists)
  - [Contribution guidelines and curation criteria](#contribution-guidelines-and-curation-criteria)
    - [Contribution guidelines](#contribution-guidelines)
    - [Curation criteria](#curation-criteria)
  - [📑 Citation](#-citation)

## Start here: the essential dozen

If you only look at twelve things, make it these:

| If you want to… | Start with | Why |
|---|---|---|
| Understand agents from first principles | [An Introduction to MultiAgent Systems](#classic-books-pre-2024) | The canonical MAS textbook |
| Connect MAS to game theory | [Shoham & Leyton-Brown](#classic-books-pre-2024) | Free online; the algorithmic/game-theoretic foundations |
| Learn MARL properly | [MARL: Foundations and Modern Approaches](#recent-books-2024) | Free modern textbook with code and slides |
| Run reproducible MARL experiments | [PettingZoo](#marl-simulation-and-classic-mas-platforms) + [BenchMARL](#marl-simulation-and-classic-mas-platforms) | Standard environment API + standardised benchmarking pipeline |
| Beat a hard cooperative benchmark | [SMACv2](#datasets-and-benchmarks) | The de-facto cooperative MARL challenge, randomised to resist overfitting |
| Evaluate social generalisation | [Melting Pot](#datasets-and-benchmarks) | Tests generalisation to new co-players, not just new tasks |
| Map the LLM multi-agent space | [Guo et al. survey](#recent-high-impact-papers-and-surveys) | The most practical overview of methods and open problems |
| Build an LLM agent team | [Framework comparison](#which-framework-should-i-use) | Pick by use case, not by star count |
| Learn from a production system | [Anthropic's multi-agent research system](#how-to-guides-and-framework-docs) | Candid engineering retrospective: when multi-agent beats single-agent |
| Know why agent teams fail | [Why Do Multi-Agent LLM Systems Fail?](#recent-high-impact-papers-and-surveys) | Empirical failure taxonomy — read before shipping |
| Connect agents to tools and data | [Model Context Protocol (MCP)](#agent-interoperability-protocols) | The industry-standard tool/context interface |
| Make agents talk across frameworks | [Agent2Agent Protocol (A2A)](#agent-interoperability-protocols) | Cross-framework agent-to-agent communication |

## Getting started: three pathways

Different readers need different entry points. Pick the pathway that matches you:

**🎓 New to MAS research** — start with Wooldridge's [An Introduction to MultiAgent Systems](#classic-books-pre-2024), then the [seminal papers](#seminal-papers-and-milestone-work) (Contract Net → BDI → Dec-POMDP complexity), and browse the [taxonomy](#field-guide-taxonomy-and-milestone-timeline) to find your subfield.

**🤖 MARL practitioner** — the [MARL book](#recent-books-2024) (free online) for foundations; [PettingZoo + BenchMARL](#marl-simulation-and-classic-mas-platforms) to run reproducible experiments; [SMACv2 and Melting Pot](#datasets-and-benchmarks) as benchmarks; MAPPO as the baseline to beat.

**🧠 LLM-agent builder** — start with the [Guo et al. survey](#recent-high-impact-papers-and-surveys) for the map, pick an orchestration framework from the [comparison table](#which-framework-should-i-use), wire up tools/communication via the [MCP and A2A protocols](#agent-interoperability-protocols), and read [Why Do Multi-Agent LLM Systems Fail?](#recent-high-impact-papers-and-surveys) before shipping.

## Explore by topic

The collection is organised by resource type below, but most readers arrive with a problem. Jump by topic:

- **Coordination and negotiation** — [seminal papers](#seminal-papers-and-milestone-work) (Contract Net, ADOPT, Max-Sum), [NegMAS](#marl-simulation-and-classic-mas-platforms), [ANAC/SCML competitions](#competitions-and-challenges).
- **Multi-agent reinforcement learning (MARL)** — [recent papers](#recent-high-impact-papers-and-surveys) (MAPPO, SMACv2, BenchMARL), [benchmarks](#datasets-and-benchmarks), [MARL platforms](#marl-simulation-and-classic-mas-platforms), the [MARL book](#recent-books-2024).
- **Emergent communication** — [DIAL and the field survey](#seminal-papers-and-milestone-work), the [EGG toolkit](#marl-simulation-and-classic-mas-platforms).
- **LLM agent teams and orchestration** — [framework comparison](#which-framework-should-i-use), [how-to guides](#how-to-guides-and-framework-docs), [recent papers](#recent-high-impact-papers-and-surveys) (AutoGen, MetaGPT, CAMEL, Generative Agents, Magentic-One).
- **Evaluation and failure modes** — [AgentBench and successors](#recent-high-impact-papers-and-surveys), [Why Do Multi-Agent LLM Systems Fail?](#recent-high-impact-papers-and-surveys), [reproducibility resources](#reproducibility-resources-and-code).
- **Safety, security, and governance** — [risk taxonomy, collusion, infectious jailbreaks, agent visibility](#recent-high-impact-papers-and-surveys), the [OWASP threat-modeling guide](#how-to-guides-and-framework-docs).
- **Simulation and robotics** — [VMAS](#marl-simulation-and-classic-mas-platforms), [Mesa/NetLogo/GAMA (ABM)](#marl-simulation-and-classic-mas-platforms), [RoboCup leagues](#competitions-and-challenges).
- **Planning and path finding (MAPF)** — [MAPF survey](#recent-high-impact-papers-and-surveys), [League of Robot Runners](#competitions-and-challenges), [MAPF community portal](#community-resources).
- **Interoperability protocols** — [MCP, A2A, AG-UI, ANP](#agent-interoperability-protocols) and the [protocol surveys](#recent-high-impact-papers-and-surveys).
- **Game theory and mechanism design** — [Shoham & Leyton-Brown](#classic-books-pre-2024), [Stanford CS 224M](#tutorials-and-courses), [OpenSpiel](#marl-simulation-and-classic-mas-platforms).

## Curated resources

Two-tier curation: **milestones** (foundational, field-shaping work) and **recent** (2021–2026, emphasising benchmarks, reproducibility, and LLM-based multi-agent systems). Entries follow the format `Title — (Year) Authors · annotation · tags`; anything not verifiable from a primary source is marked **unspecified**. Star counts are live badges via [shields.io](https://shields.io/).

### Classic books (pre-2024)

- [**An Introduction to MultiAgent Systems** (2nd ed.)](https://www.cs.ox.ac.uk/people/michael.wooldridge/pubs/imas//IMAS2e.html) - (2009) by Michael Wooldridge
  A widely used MAS textbook covering agent concepts, interaction, coordination, and foundational theory; a solid "first principles" entry point for the broader MAS canon beyond MARL. `foundations`, `agents`, `coordination`.

- [**Multiagent Systems: Algorithmic, Game-Theoretic, and Logical Foundations**](http://www.masfoundations.org/) - (2009) by Yoav Shoham, Kevin Leyton-Brown
  Core reference connecting MAS with game theory, mechanism design, and computational foundations; also supported by a freely accessible "rough version" linked from a canonical course page. `game-theory`, `mechanism-design`, `foundations`.

- [**Multi-Agent Systems: A Modern Approach to Distributed Artificial Intelligence**](https://mitpress.mit.edu/9780262731317/multi-agent-systems/) - (1999) by Gerhard Weiss (editor)
  Classic edited volume spanning early MAS themes (coordination, communication, architectures) from distributed AI roots; useful for historical depth and breadth. `distributed-ai`, `foundations`, `architectures`.


### Recent books (2024–)

- [**Multi-Agent Systems: A Contemporary Treatment**](https://books.bloo-mind.ai/masact/) - (2026) by Dell Zhang, Jun Wang, Benjamin Chang
  A working draft textbook teaching the concepts and techniques of multi-agent systems in the era of LLMs; freely readable online. *Maintainer-affiliated* — see the [editorial policy](#editorial-policy). `textbook`, `LLM-agents`, `foundations`.

- [**Agents and Multi-Agent Systems Development: Platforms, Toolkits, Technologies**](https://link.springer.com/book/10.1007/978-3-032-01082-7) - (2026) by R. Collier, V. Mascardi, A. Ricci (editors)
  A snapshot of the current state of the art in tools, frameworks, and techniques for designing and implementing multi-agent systems; includes a chapter on "Agent Toolkits Anno 2025." `MAS-engineering`, `platforms`, `toolkits`.

- [**Design Multi-Agent AI Systems Using MCP and A2A**](https://www.packtpub.com/en-sg/product/design-multi-agent-ai-systems-using-mcp-and-a2a-9781806116461) - (2026) by Gigi Sayfan
  Hands-on guide to building a production-ready multi-agent AI framework from scratch in Python; covers tool use, memory via MCP, collaborative agent workflows with A2A, observability, and human-in-the-loop patterns. Companion code on GitHub. `LLM-agents`, `MCP`, `A2A`, `practice`.

- [**Agentic AI: Theories and Practices**](https://link.springer.com/book/10.1007/978-3-031-90026-6) - (2025) by Ken Huang (editor)
  Analyses the rise of generative AI agents (agentic AI) across industries, covering development, applications, and implications from finance to healthcare. `agentic-ai`, `LLM-agents`, `applications`.

- [**AI Agents in Action**](https://www.manning.com/books/ai-agents-in-action) - (2025) by Micheal Lanham
  Practitioner guide to building multi-agent AI systems using modern frameworks (LangChain, AutoGen, CrewAI); covers knowledge management, memory systems, and collaborative multi-agent architectures. `LLM-agents`, `multi-agent`, `practice`.

- [**Building Applications with AI Agents**](https://www.oreilly.com/library/view/building-applications-with/9781098176495/) - (2025) by Michael Albada
  A practical, research-based approach to designing and implementing single- and multi-agent systems, covering coordination techniques and communication methods for agent systems. `LLM-agents`, `multi-agent`, `practice`.

- [**Designing Multi-Agent Systems: Principles, Patterns, and Implementation for AI Agents**](https://multiagentbook.com/) - (2025) by Victor Dibia
  A first-principles guide to designing multi-agent applications, walking through building a feature-complete framework from scratch; by a core AutoGen contributor at Microsoft Research. Companion code on GitHub. `LLM-agents`, `multi-agent`, `practice`.

- [**Multi-Agent Reinforcement Learning: Foundations and Modern Approaches**](https://www.marl-book.com/) - (2024) by Stefano V. Albrecht, Filippos Christianos, Lukas Schäfer
  A modern MARL textbook focusing on models, solution concepts, algorithms, and practical challenges; associated with a companion website and learning materials (slides/code). `MARL`, `RL`, `game-theory`, `reproducibility`, `code`.


### Tutorials and courses

- [**European Agent Systems Summer School (EASSS)**](https://easss.upb.ro/) - (2025) by EURAMAS community
  Long-running summer school (since 1999) offering introductory and advanced courses across autonomous agents and MAS, aimed at researchers and students. `community`, `tutorials`, `foundations`.

- [**AAMAS tutorials programme**](https://aamas2025.org/index.php/conference/program/tutorials/) - (2025) by AAMAS organisers
  Tutorials highlight evolving MAS topics; the official programme page provides titles/abstracts and can seed curated "learning pathways" each year. `community`, `tutorials`, `MAS`.

- [**Cooperative AI Summer School**](https://www.cooperativeai.com/summer-school/summer-school-2025) - (2025) by Cooperative AI community
  A summer school aimed at grounding participants in cooperative AI (overlapping with MAS/MARL, incentives, and human/agent cooperation). `cooperative-ai`, `MARL`, `incentives`.

- [AAMAS 2025 tutorial: **RL in Automated Negotiation (T1)**](https://github.com/yasserfarouk/aamas2025rlneg) - (2025) by Yasser Farouk
  A concrete tutorial artefact (slides/code) framing negotiation as a multi-agent RL problem; useful for bridging MAS negotiation and learning-based approaches. `negotiation`, `MARL`, `tutorial`.

- [Stanford **CS 224M: Multi Agent Systems**](https://web.stanford.edu/class/cs224m/) - (2014) by Stanford course staff; Instructor: Yoav Shoham
  A game-theory-and-mechanism-design-heavy MAS course page with structured readings, lecture materials via edX links, and a direct tie-in to the Shoham & Leyton-Brown textbook. `game-theory`, `mechanism-design`, `foundations`.


### How-to guides and framework docs

- [**How we built our multi-agent research system**](https://www.anthropic.com/engineering/built-multi-agent-research-system) - (2025) by Anthropic
  A widely read engineering retrospective on building a production orchestrator–worker multi-agent system: when multi-agent beats single-agent, prompt/tool design for delegation, and evaluation lessons. `LLM-agents`, `orchestration`, `engineering`.

- [**Multi-Agentic System Threat Modeling Guide**](https://genai.owasp.org/resource/multi-agentic-system-threat-modeling-guide-v1-0/) - (2025) by OWASP Gen AI Security Project
  Applies OWASP's agentic-AI threats-and-mitigations taxonomy to concrete multi-agent deployments through worked threat models; the current practitioner reference for securing multi-agent LLM systems. `security`, `threat-modeling`, `how-to`.

- [**LangGraph documentation

…

## Source & license

This open-source MCP server is cataloged on AgentStack and links to its original source — we do not rehost the code.

- **Author:** [bloo-mind](https://github.com/bloo-mind)
- **Source:** [bloo-mind/awesome-multi-agent-systems](https://github.com/bloo-mind/awesome-multi-agent-systems)
- **License:** MIT

Install and usage instructions live in the source repository linked above.

## Pricing

- **Free** — Free

## Security capabilities

Automated source analysis of v0.1.0 — what this tool can access:

- **Network access:** no
- **Filesystem access:** no
- **Shell / process execution:** no
- **Environment & secrets:** no
- **Dynamic code execution:** no

*"Yes" means the capability is present in the source — more access means more to trust, not that it is unsafe.*


## Versions

- **0.1.0** — security scan: passed — Imported from the upstream source.

## Links

- Listing page: https://agentstack.voostack.com/l/mcp-bloo-mind-awesome-multi-agent-systems
- Seller: https://agentstack.voostack.com/s/bloo-mind
- Browse the marketplace: https://agentstack.voostack.com/browse

---
Listed on AgentStack — the marketplace for AI agent skills and MCP servers. Every listing is security-reviewed. Creators keep 70%.