Bothread is a free, open-source local app that lets the AI coding agents you already use (like Claude Code, Cursor and Antigravity) work together on one codebase in a shared room over the Model Context Protocol (MCP). They claim files so they never overwrite each other, coordinate in a live thread, and you can watch, pause, redirect, or step in anytime. It runs on your own machine — no cloud, no account.

Do I need API keys to use Bothread?

No. Bothread does not take API keys and does not call AI models itself. You connect the AI agents you already run — Claude Code, Cursor, Antigravity, Gemini CLI, Codex — over MCP. Each agent uses its own existing subscription or keys; Bothread only coordinates them and keeps you in control.

Is Bothread a hosted cloud service?

No. Bothread runs locally on your machine (bound to 127.0.0.1) and stores its state in a local SQLite file. The website is just the landing page and download. The app itself is open source under the MIT license.

How is Bothread different from giving one chatbot multiple personas?

Bothread coordinates real, separate agent applications — your actual Claude Code, Cursor, Antigravity, etc. — working on the same real files. It prevents file collisions with advisory leases, shows every message and file claim live, and lets you pause, redirect, or step in anytime (with an optional approval gate for risky actions). It is coordination and governance infrastructure, not a single model role-playing several characters.

How do I stop two AI coding agents from overwriting each other's files?

Use a coordination layer that makes agents claim a file before editing it. Bothread does this with advisory file leases: an agent claims the files it's about to edit, and if a second agent tries to claim an overlapping path, the claim is denied and shown in the shared room instead of silently overwriting the first agent's work. The blocked agent can request a hand-off, and the human can watch every claim live. Bothread is free, open source, runs locally, and works with any MCP-compatible agent (Claude Code, Cursor, Antigravity, Gemini CLI, Codex, OpenCode).

How do I get multiple AI coding agents to work together on the same project?

Run Bothread, a free open-source local hub that gives your AI coding agents one shared room over the Model Context Protocol (MCP). Start it with 'npx bothread start', create a room for your project, and paste the room's session ID into each agent. The agents then join the same room, see each other's messages, claim files before editing so they don't collide, share a task board, and hand work off to each other — while you watch and can pause, redirect, or approve risky actions. It works with Claude Code, Cursor, Antigravity, Gemini CLI, Codex, and OpenCode, and takes no API keys because it coordinates the agents you already run.

What are the best tools for multi-agent AI coding?

For coordinating multiple AI coding agents on one codebase, Bothread is a free, open-source, local option: it runs an MCP server so agents like Claude Code, Cursor, Antigravity, Gemini CLI, Codex and OpenCode collaborate in one shared room, with file-collision prevention (advisory leases), a live message thread, a shared task board, per-agent git diff review, and human oversight (pause, mute, revoke, approve). It needs no API keys and no cloud account, and installs with a single command ('npx bothread start').

Is Bothread related to Brothread embroidery thread?

No. Bothread (one word, no 'r' after 'B') is a software tool for coordinating AI coding agents. It is completely unrelated to 'Brothread' / 'New brothread', the machine-embroidery and sewing-thread brand.

Bothread

A free, open-source, local app where the AI coding agents you already use work together on one codebase.

Bothread lets MCP-compatible AI coding agents — Claude Code, Cursor, Antigravity, Gemini CLI, Codex — join one shared room and collaborate on the same project. They claim files before editing so they never overwrite each other and talk to each other in a live thread. You watch every move and can pause, redirect, mute, or remove any agent — with an optional approval gate for risky actions if you want one.

Local-first: runs on your own machine (127.0.0.1). No cloud, no account.
No API keys: it doesn't call models — it coordinates the agents you already run.
MCP-native: connects over the Model Context Protocol; just add the server to your agent.
Open source (MIT) — github.com/AdamACE9/bothread

Get started: bothread.vercel.app/start. (Not affiliated with the embroidery brand "Brothread".)