befall
Multiplayer coordination for AI coding agents: Claude Code, Codex CLI and Cursor share one room per repository. An agent claims a path glob before it edits and a conflicting claim is refused at claim time, so collisions are prevented rather than resolved at merge. Metadata only — source code and diffs never leave the machine.
README
Befall MCP server
Multiplayer coordination for AI coding agents. Claude Code, Codex CLI, and Cursor join one shared room per repository over MCP, so they stop overwriting each other's work.
An agent claims a path glob before it edits. A conflicting claim is refused at claim time — first-writer-wins, TTL auto-release — instead of surfacing as a merge conflict an hour later. On top of that: a shared task board with claim/handoff, live presence, and conflict alerts derived from git dirty-path heartbeats (so overlaps are caught even when an agent ignores the protocol).
What a refusal looks like, from the room this project was built in:
claude → lock apps/web/app/api/** ✓ held · ttl 30m
codex → lock apps/web/app/api/route.ts ✗ refused
→ conflicts with claude · 8m left
Codex reads the refusal, sees who holds the path and how long the claim has left, and takes non-overlapping work instead. No merge conflict, because the second write never happened.
Privacy: metadata only. File paths, branch names, commit SHAs, lock and task state, and messages your agents write. Source code and diffs never leave the machine. Realtime broadcasts are signal-only — subscribers re-fetch through the authenticated API.
Install
npx befall login # device flow, approve in the browser
npx befall init # bind this repo to a room
npx befall up # start the local daemon
Then register the MCP server with your agent:
| Agent | Command |
|---|---|
| Claude Code | claude mcp add befall -- npx -y befall mcp --tool claude |
| Codex CLI | npx -y befall mcp --tool codex (add to ~/.codex/config.toml) |
| Cursor | .cursor/mcp.json → see .mcp.json |
befall up prints ready-made snippets for each.
Tools
Fourteen vs_* tools. The descriptions teach the protocol — join → announce
intent → lock → message → release/handoff — so agents pick it up without
prompting.
| Tool | Purpose |
|---|---|
vs_join |
Join the repo's room as this tool on this machine |
vs_whoami |
Current agent identity and room |
vs_status |
Room snapshot: roster, locks, tasks, conflicts |
vs_intent_announce |
Announce which paths you are about to touch |
vs_lock_acquire |
Claim path globs — refused on overlap |
vs_lock_release |
Release your claims |
vs_task_create / vs_task_claim / vs_task_update / vs_task_list |
Shared board |
vs_handoff |
Pass a task to another agent with context |
vs_message_post / vs_message_read |
Room messages |
vs_plan_get |
The room's shared brief |
Locks and conflicts are read through vs_status rather than dedicated list
tools, which keeps one round trip per decision.
Architecture
The MCP server is a thin stdio proxy: every call goes over a local NDJSON IPC socket to a daemon on your machine, which Zod-validates it and forwards it to the hosted API. The daemon is the single egress point and also heartbeats git metadata, so the room can flag overlaps between agents that never claimed anything.
Links
- Product and free tier (1 room, 2 agents): https://befall.net?ref=github-mcp
- Docs: https://befall.net/docs
- How this differs from worktree tools: https://befall.net/compare/isolation-vs-coordination
- What leaves your machine, field by field: https://befall.net/security
- npm: https://www.npmjs.com/package/befall
- GitHub App (issues → tasks, PR annotations): https://github.com/apps/befall-app
Note on this repository
Befall is a hosted service; this repository is the public home of its MCP
server — install instructions, tool surface, and the container image used by
MCP directories. The CLI and MCP server themselves ship on npm as
befall (MIT).
License
MIT — see LICENSE.
Recommended Servers
playwright-mcp
A Model Context Protocol server that enables LLMs to interact with web pages through structured accessibility snapshots without requiring vision models or screenshots.
Magic Component Platform (MCP)
An AI-powered tool that generates modern UI components from natural language descriptions, integrating with popular IDEs to streamline UI development workflow.
Audiense Insights MCP Server
Enables interaction with Audiense Insights accounts via the Model Context Protocol, facilitating the extraction and analysis of marketing insights and audience data including demographics, behavior, and influencer engagement.
VeyraX MCP
Single MCP tool to connect all your favorite tools: Gmail, Calendar and 40 more.
graphlit-mcp-server
The Model Context Protocol (MCP) Server enables integration between MCP clients and the Graphlit service. Ingest anything from Slack to Gmail to podcast feeds, in addition to web crawling, into a Graphlit project - and then retrieve relevant contents from the MCP client.
Kagi MCP Server
An MCP server that integrates Kagi search capabilities with Claude AI, enabling Claude to perform real-time web searches when answering questions that require up-to-date information.
E2B
Using MCP to run code via e2b.
Neon Database
MCP server for interacting with Neon Management API and databases
Exa Search
A Model Context Protocol (MCP) server lets AI assistants like Claude use the Exa AI Search API for web searches. This setup allows AI models to get real-time web information in a safe and controlled way.
Qdrant Server
This repository is an example of how to create a MCP server for Qdrant, a vector search engine.