phuryn

phuryn / swarm-protocol

Public

Coordination protocol for agent-first teams. No UI. No sprints. No Jira. Just state sync.

agent-coordination ai-coding claude-code codex cursor
22
4
100% credibility
Found Mar 14, 2026 at 20 stars -- GitGems finds repos before they trend. Get early access to the next one.
Sign Up Free
AI Analysis
TypeScript
AI Summary

Swarm Protocol provides a shared state synchronization system for AI agent teams to coordinate tasks, detect file conflicts, and hand off work on shared codebases without user interfaces or meetings.

How It Works

1
🔍 Discover coordination magic

You learn about a simple system that lets your team's AI helpers work together on projects without endless meetings or messages.

2
🚀 Start the shared hub

You launch the central spot where all AI helpers can check in and stay synced with a quick setup.

3
👥 Build your team space

You create a team area and list out goals like fixing a feature or adding a new page.

4
📝 Drop in the secret note

You add a short instruction file to your project folder, and your AI helpers instantly know how to team up.

5
🤖 Helpers claim their tasks

AI buddies grab open goals, check for file overlaps, and ping if someone's already working there.

6
💓 Keep the pulse going

They send quick updates while working, finish up, and automatically free the next goals for others.

🎉 Projects flow perfectly

Your team builds software faster with no clashes, lost context, or waiting – just smooth, happy progress.

Sign up to see the full architecture

5 more

Sign Up Free

Star Growth

See how this repo grew from 20 to 22 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 swarm-protocol?

Swarm-protocol is a headless coordination protocol for agent-first dev teams, syncing state so AI agents like Claude Code can claim work, detect file conflicts, send heartbeats, and hand off tasks without Slack or Jira. Built in TypeScript with Node.js and PostgreSQL, it exposes 19 MCP tools—like check_conflicts for file overlaps and get_context for full task handoff—turning intents into a lifecycle of draft, open, claimed, done, with auto-unblocking dependencies. Drop a markdown snippet into your repo's CLAUDE.md, spin up the Dockerized Postgres server, and agents coordinate across sessions on a shared codebase.

Why is it gaining traction?

In a sea of coordination protocols and LLM coordination tools on GitHub, it stands out with zero-config integration for Claude Code, advisory conflict checks before edits, and context packages that bundle dependencies, signals, and team norms in one call. Devs hook into its swarm communication protocol via stdio MCP servers—no REST, no UI—for seamless inter-human, inter-session sync that existing single-player agent tools ignore. Early adopters praise the primitives: intents track outcomes, claims prevent merges, signals notify blocks.

Who should use this?

Multi-dev teams where each runs AI agents (Claude Code, etc.) as the main coding interface on one repo—think remote backend squads avoiding duplicate PRs or stalled handoffs. Pairs or small groups scaling agent workflows past solo sessions, especially if you're already using MCP-compatible tools and hate status pings.

Verdict

Alpha with solid docs, integration tests, and MIT license, but 17 stars and 1.0% credibility score signal high risk for production—fork and co-build if agent teams are your future. Worth a Docker spin-up for multiplayer LLM coordination experiments today.

(198 words)

Sign up to read the full AI review Sign Up Free

Similar repos coming soon.