opencode-chatgpt-bridge
A local MCP bridge that lets ChatGPT control opencode sessions for code modification, file reading, and repository management on your own computer.
README
opencode-chatgpt-bridge
A standalone local MCP bridge that lets ChatGPT control opencode sessions running on your own computer.
ChatGPT / ChatGPT Mobile
-> HTTPS tunnel
-> opencode-chatgpt-bridge /mcp
-> local opencode serve
-> local repository
This project is intentionally separate from LocalAnt. It is focused only on the ChatGPT <-> opencode bridge use case.
Quick start
git clone https://github.com/yuga-hashimoto/opencode-chatgpt-bridge.git
cd opencode-chatgpt-bridge
pnpm install
pnpm run build
pnpm run init -- --allowed-roots /Volumes/MOVESPEED/Documents/GitHub
pnpm start
init creates a ready-to-use .env with a random bridge token, your allowed repo roots, automatic port fallback, and Tailscale Funnel enabled by default.
When the bridge starts, it prints a setup guide with:
- local health URL
- local MCP URL
- automatic fallback port when the preferred port is already in use
- public HTTPS MCP URL when tunnel is enabled; Tailscale background service defaults to port 10000 to avoid LocalAnt using 443 and other active Funnel listeners
- ChatGPT settings link
- connector name, description, and URL to paste
- header auth and URL-token fallback
- opencode CLI status and setup notes
- first ChatGPT prompt to try
Requirements
- Node.js 20+
- pnpm 10+
opencodeinstalled and authenticated locallycloudflaredfor the default tunnel flow, or your own HTTPS tunnel
Check opencode:
opencode --version
opencode
Run opencode once inside a repo and make sure your model provider is configured before expecting ChatGPT to drive it. The bridge starts opencode serve for each repo and talks to it over HTTP. After creating a bridge session, call opencode_capabilities from ChatGPT to inspect connected providers, available auth methods, and default model configuration.
Commands
opencode-chatgpt-bridge init --allowed-roots /path/to/repos
opencode-chatgpt-bridge start
opencode-chatgpt-bridge doctor
From source, use pnpm:
pnpm run init -- --allowed-roots /path/to/repos
pnpm start
pnpm run doctor
Background mode on macOS
You do not need to keep a terminal open. Install the bridge as a user LaunchAgent:
pnpm run build
pnpm run install-service
pnpm run service-status
Logs are written to:
~/.opencode-chatgpt-bridge/bridge.log
~/.opencode-chatgpt-bridge/bridge.err.log
To stop background mode:
pnpm run uninstall-service
The service uses this repository directory as its working directory, loads .env, starts the bridge, and keeps it alive after login.
ChatGPT setup
Open ChatGPT Web and go to:
https://chatgpt.com/#settings/Connectors
Manual path:
Settings -> Apps & Connectors -> Advanced settings -> enable Developer mode
Settings -> Connectors -> Create
Use the values printed by the bridge. They look like this:
Connector name: opencode local bridge
Description: Control local opencode sessions, inspect diffs, and manage local coding tasks.
Connector URL: https://example.trycloudflare.com/mcp?token=<generated-token>
If your ChatGPT connector UI supports auth headers, you can use the plain /mcp URL and set:
Authorization: Bearer <OPENCODE_BRIDGE_TOKEN>
If not, use the printed ?token= URL fallback.
Once linked on ChatGPT Web, the connector should be available in ChatGPT mobile apps as well.
opencode setup
No extra opencode project configuration is required by the bridge, but opencode itself must be usable locally.
The bridge starts opencode like this:
OPENCODE_SERVER_USERNAME=opencode \
OPENCODE_SERVER_PASSWORD=<generated-or-env-password> \
opencode serve --hostname 127.0.0.1 --port <auto>
Notes:
- If
OPENCODE_SERVER_PASSWORDis not set, the bridge generates a random password per managed opencode server. - If
OPENCODE_BASE_URLis set, the bridge uses that existing opencode server instead of spawning one. - The opencode server is kept on
127.0.0.1; only the bridge is exposed to ChatGPT. - Provider/model login is handled by opencode. Run
opencodein a terminal first and confirm it can answer/edit before using ChatGPT.
Features
- Streamable HTTP MCP endpoint at
/mcpfor ChatGPT connectors and MCP clients. - Per-repository
opencode serveprocess management. - Safe repository allow-listing with realpath checks.
- Bridge bearer token for exposed/tunneled use.
- URL-token fallback for connector UIs that do not support custom headers.
- Persistent bridge session mapping in
~/.opencode-chatgpt-bridge/sessions.json. - Tools for session creation, async prompts, polling, message retrieval, diff review, abort, permission responses, file reads, file search, VCS status, agents, slash commands, and provider/model diagnostics.
- Optional Cloudflare quick tunnel launcher.
- TypeScript, strict typecheck, and unit tests.
MCP tools
Bridge and project tools
bridge_health- inspect bridge config and managed opencode processes.list_projects- list Git repos under the allowed roots.
opencode process/session tools
opencode_start- start or reuse anopencode serveprocess for a repo.opencode_stop- stop one or all managed opencode servers.opencode_create_session- create a new opencode session and return abridgeSessionId.opencode_list_sessions- list bridge sessions known to this bridge.opencode_get_session_status- poll status for a session.opencode_send_message- send a prompt; defaults to async mode.opencode_get_messages- fetch session transcript/messages.opencode_get_diff- fetch file diffs for a session.opencode_abort- abort a running session.opencode_respond_permission- respond to opencode permission prompts.
project inspection tools
opencode_read_file- read a file through opencode.opencode_find_files- fuzzy-find files through opencode.opencode_vcs_status- get VCS and file status.opencode_capabilities- list opencode agents, slash commands, providers, auth methods, and default model config.
Example ChatGPT prompt
Use opencode local bridge. First call bridge_health and list_projects.
Then create a session for /Volumes/MOVESPEED/Documents/GitHub/my-repo,
ask opencode to fix the README, poll status, and show opencode_get_diff.
Security model
This bridge can cause local code modifications through opencode. Treat it as a local automation gateway.
Default protections:
opencodeitself is bound to127.0.0.1.- Repositories must be under
OPENCODE_BRIDGE_ALLOWED_ROOTS. - The bridge supports bearer-token authentication and URL-token fallback.
- No arbitrary shell execution tool is exposed by this bridge.
- Diffs are first-class so clients can inspect changes before committing.
Strongly recommended:
- Always set
OPENCODE_BRIDGE_TOKENbefore exposing through any tunnel. - Do not bind the bridge to
0.0.0.0unless you know exactly what network can reach it. - Keep
OPENCODE_BRIDGE_ALLOWED_ROOTSnarrow. - Review
opencode_get_diffbefore committing or pushing generated changes.
Development
pnpm install
pnpm run typecheck
pnpm test
pnpm run build
pnpm run validate
License
MIT
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.