balajivis

balajivis / sutra-mas

Public

36,299 multi-agent systems papers collected, 17,969 analyzed with coordination patterns, embeddings, and a 16-cluster taxonomy — the largest structured MAS corpus bridging 30 years of classical research with modern LLM agents

coordination embeddings llm-agents mas multi-agent-systems
23
3
100% credibility
Found Feb 17, 2026 at 15 stars -- GitGems finds repos before they trend. Get early access to the next one.
Sign Up Free
AI Analysis
Python
AI Summary

Sutra provides the world's largest organized collection of multi-agent systems research papers with analysis, a 16-cluster taxonomy, and experiments testing classical coordination patterns on modern AI agent tasks.

How It Works

1
🔍 Discover Sutra

You find Sutra, a huge collection of research on how teams of AI helpers work together over 30 years.

2
📥 Grab the ready-to-use files

Download simple files like spreadsheets and lists of papers right from the collection—no setup needed.

3
📖 Browse your new knowledge library

Open the files to see thousands of papers sorted into 16 easy categories like team planning or shared ideas.

4
🏛️ Dive into a category

Pick one group, like 'team structures,' and read summaries of key papers from old classics to newest ideas.

5
🧪 Test old teamwork tricks

Try simple tests comparing different ways AI helpers coordinate, seeing which work best today.

🎉 Build better AI teams

Use the insights to make your own reliable group of AI assistants that don't fail like others do.

Sign up to see the full architecture

4 more

Sign Up Free

Star Growth

See how this repo grew from 15 to 23 stars Sign Up Free
Repurpose This Repo

Repurpose is a Pro feature

Generate ready-to-use prompts for X threads, LinkedIn posts, blog posts, YouTube scripts, and more -- with full repo context baked in.

Unlock Repurpose
AI-Generated Review

What is sutra-mas?

Sutra-mas is a Python toolkit delivering the largest structured corpus of multi-agent systems research: 36,299 papers collected, 17,969 analyzed with coordination patterns, embeddings, and a 16-cluster taxonomy bridging classical MAS to modern LLM agents. Developers get ready-to-load datasets like JSONL exports for querying papers by pattern (blackboard, contract net) or cluster, plus a citation graph and lost-canary classics ignored by today's LLM work. Run the experiment harness to test 11 coordination patterns across 5 benchmarks, comparing baselines to see classical designs outperform naive multi-agent setups.

Why is it gaining traction?

It stands out by reconnecting forgotten MAS coordination protocols—like blackboard control shells scoring 95/100 on code review—with LLM agents, backed by empirical results in 58 reproducible runs. Unlike raw paper lists, you query "modern blackboard papers missing Nii 1986" or extend patterns with your own agents. The hook: instant baselines proving 50% token savings and error prevention without prompt hacks.

Who should use this?

AI engineers building LLM agent teams for code review, planning, or research synthesis, tired of cascading failures in naive multi-agent flows. MAS researchers validating classical patterns on LLM benchmarks. Framework authors like LangGraph users seeking taxonomy-grounded coordination baselines.

Verdict

Grab the data exports and harness for agent prototyping—massive corpus value despite 14 stars and 1.0% credibility signaling early maturity. Solid docs and Apache 2.0 license make it extensible; run experiments first to confirm fit.

(198 words)

Sign up to read the full AI review Sign Up Free

Similar repos coming soon.