Codex Claude Bridge

Claude Code and Codex CLI talk to each other through one markdown file in your project. No server, no hooks, no plugin.

1# Codex Claude Bridge

README.mdMCP server

Transport: stdio
Status: Local

Power your agent

Add this server

Install from the publisher. Not a PlugMCP endpoint.

Run on PlugMCP

Host a copy with our keys and logs.

No packaged release in the registry yet. Follow the repository README.

RepositoryPublisher installStep-by-step: add this in Claude

Remote URL or a command on your machine? How the two installs differ

Run on PlugMCP

Host a copy with our keys and logs.

A new server you create. It is not this listing's official endpoint. Use the publisher install when you want their URL and their credentials.

Launch

Server facts

Local

Transport: stdio
Last checked: 2026-10-26

Stars: 62
Tools: —
Uptime: —

Tools

No tool list yet. Remote servers get one after the first health check.

FAQ

Questions

How do I install Codex Claude Bridge?

Copy the stdio config for Cursor, Claude Desktop, Claude Code, or VS Code on this page. It uses the publisher's install, not a PlugMCP proxy.

Does Codex Claude Bridge need an API key?

No required environment variable is listed on this record.

What transport does Codex Claude Bridge use?

Codex Claude Bridge runs locally over stdio.

When was Codex Claude Bridge last checked?

Last checked 2026-10-26. Claude Code and Codex CLI talk to each other through one markdown file in your project. No server, no hooks, no plugin.

From the publisher

Documentation

Codex Bridge

Claude Code and Codex CLI talk to each other through one markdown file in your project. No server, no hooks, no plugin. You read along.

Claude Code and Codex CLI take turns appending messages to talkto.md. After posting, each agent runs a watch command that wakes it when the other agent's reply is complete.

Quick start

Add talkto.md to your project root:

curl -fsSL https://raw.githubusercontent.com/abhishekgahlot2/codex-claude-bridge/main/talkto.md -o talkto.md

Open claude and codex in that project, or in any of its git worktrees.

Tell one of them what to discuss:

discuss redis vs memcached with codex via talkto.md, keep going until you agree

Tell the other one check talkto.md. From then on they keep going on their own.

How it works

The rules sit at the top of talkto.md, so each agent learns the protocol by reading the file. There is nothing to install.

An agent appends a message that ends with Waiting on: <who> — <what>.
It runs the watch command from the file in the background.
- Claude Code runs it with Monitor and is woken when it prints.
- Codex CLI runs it in a background terminal and keeps polling, so its turn stays open.
The watch exits when a complete message addressed to that agent appears. The agent reads it, replies, and waits again.
Waiting on: nobody — done ends the conversation.

The watch compares against the agent's own last message, not a line count. A reply that lands before the watch starts is still caught, and a half-written one waits until its Waiting on: line exists. Every git worktree resolves to the main checkout's talkto.md.

Message format

## Codex → Claude — 2026-10-08 15:55 IST · round 1
Redis by default. Memcached only for a huge, multi-threaded cache of small blobs.

Waiting on: Claude — counterpoint or agreement

## Claude → Codex — 2026-10-08 15:57 IST · round 2
Reply to: Codex → Claude — 2026-10-08 15:55 IST · round 1
Agreed. Confirm in one line and close.

Waiting on: Codex — confirm and close

Part	Rule
Heading	`## <From> → <To> — <date time zone> · <topic>`. The watch looks for the arrow.
`Reply to:`	Optional. The heading you are answering.
Last line	`Waiting on: <who> — <what>`. Hands over the turn. A message counts as sent only once this line exists.
Status	One block at the top of the file, edited in place: whose turn it is.
Append-only	Nobody edits another agent's message.

Optional: skip naming the file

Add one line to AGENTS.md, which Codex reads, and to CLAUDE.md, which Claude reads:

To talk to the other agent (Claude or Codex), use talkto.md at the repo root and follow the rules at its top.

Then discuss this with codex is enough.

Limits

An agent waits only while its turn is running. If Esc, an error or a usage limit cuts a turn short, say check talkto.md.
Both agents must run on the same machine.
Anything appended to talkto.md reaches both agents as a message.
Agents read only the latest messages, but a long file still costs context. Start a fresh file when a topic is done.

Tested

Test	Result
Codex CLI 0.161, live	From one plain prompt: posted, ran the watch in a background terminal, waited through a 100-second gap, and replied 39 seconds after Claude wrote. Checked 6 times, 30 to 50 seconds each.
Claude Code with Sonnet 5, live	From one plain prompt: posted, ran the watch with Monitor, and replied 59 and then 10 seconds after each Codex message, the second of which came 90 seconds later. Closed with `Waiting on: nobody — done`.
Watch command, bash and zsh	Reply before the watch starts, half-written reply, bold `Waiting on:` line, message to someone else (ignored), worktree subfolder, repo subfolder, folder outside git.

Why not A2A or MCP

There is no standard for agents sharing a conversation file. A2A is a JSON-RPC protocol between agent servers, and neither CLI speaks it. MCP gives one agent tools, not a conversation with another live session. A markdown file needs nothing from either CLI: you can read it, and both agents read and write it with tools they already have.

Versions

Version	Design
v0.3.0	One markdown file. The rules and the watch command live in the file.
v0.2.0	Markdown file, Stop hooks, and a Claude channel. Codex waited inside its Stop hook.
v0.1.0	Blocking Codex MCP tool, Claude channel, and an HTTP server with a web UI.

License

MIT

Codex Claude Bridge

abhishekgahlot262starsClaim this server

Claude Code and Codex CLI talk to each other through one markdown file in your project. No server, no hooks, no plugin.

Install Run on PlugMCPClaim this server

GitHub

Run on PlugMCP

1# Codex Claude Bridge

README.mdMCP server

Transport: stdio
Status: Local

Power your agent

Add this server

Install from the publisher. Not a PlugMCP endpoint.

Run on PlugMCP

Host a copy with our keys and logs.

No packaged release in the registry yet. Follow the repository README.

RepositoryPublisher installStep-by-step: add this in Claude

Remote URL or a command on your machine? How the two installs differ

Run on PlugMCP

Host a copy with our keys and logs.

A new server you create. It is not this listing's official endpoint. Use the publisher install when you want their URL and their credentials.

Launch

Server facts

Local

Transport: stdio
Last checked: 2026-10-26

Stars: 62
Tools: —
Uptime: —

Tools

No tool list yet. Remote servers get one after the first health check.

FAQ

Questions

How do I install Codex Claude Bridge?

Copy the stdio config for Cursor, Claude Desktop, Claude Code, or VS Code on this page. It uses the publisher's install, not a PlugMCP proxy.

Does Codex Claude Bridge need an API key?

No required environment variable is listed on this record.

What transport does Codex Claude Bridge use?

Codex Claude Bridge runs locally over stdio.

When was Codex Claude Bridge last checked?

Last checked 2026-10-26. Claude Code and Codex CLI talk to each other through one markdown file in your project. No server, no hooks, no plugin.

From the publisher

Documentation

Codex Bridge

Claude Code and Codex CLI talk to each other through one markdown file in your project. No server, no hooks, no plugin. You read along.

Claude Code and Codex CLI take turns appending messages to talkto.md. After posting, each agent runs a watch command that wakes it when the other agent's reply is complete.

Quick start

Add talkto.md to your project root:

curl -fsSL https://raw.githubusercontent.com/abhishekgahlot2/codex-claude-bridge/main/talkto.md -o talkto.md

Open claude and codex in that project, or in any of its git worktrees.

Tell one of them what to discuss:

discuss redis vs memcached with codex via talkto.md, keep going until you agree

Tell the other one check talkto.md. From then on they keep going on their own.

How it works

The rules sit at the top of talkto.md, so each agent learns the protocol by reading the file. There is nothing to install.

An agent appends a message that ends with Waiting on: <who> — <what>.
It runs the watch command from the file in the background.
- Claude Code runs it with Monitor and is woken when it prints.
- Codex CLI runs it in a background terminal and keeps polling, so its turn stays open.
The watch exits when a complete message addressed to that agent appears. The agent reads it, replies, and waits again.
Waiting on: nobody — done ends the conversation.

Message format

## Codex → Claude — 2026-10-08 15:55 IST · round 1
Redis by default. Memcached only for a huge, multi-threaded cache of small blobs.

Waiting on: Claude — counterpoint or agreement

## Claude → Codex — 2026-10-08 15:57 IST · round 2
Reply to: Codex → Claude — 2026-10-08 15:55 IST · round 1
Agreed. Confirm in one line and close.

Waiting on: Codex — confirm and close

Part	Rule
Heading	`## <From> → <To> — <date time zone> · <topic>`. The watch looks for the arrow.
`Reply to:`	Optional. The heading you are answering.
Last line	`Waiting on: <who> — <what>`. Hands over the turn. A message counts as sent only once this line exists.
Status	One block at the top of the file, edited in place: whose turn it is.
Append-only	Nobody edits another agent's message.

Optional: skip naming the file

Add one line to AGENTS.md, which Codex reads, and to CLAUDE.md, which Claude reads:

To talk to the other agent (Claude or Codex), use talkto.md at the repo root and follow the rules at its top.

Then discuss this with codex is enough.

Limits

An agent waits only while its turn is running. If Esc, an error or a usage limit cuts a turn short, say check talkto.md.
Both agents must run on the same machine.
Anything appended to talkto.md reaches both agents as a message.
Agents read only the latest messages, but a long file still costs context. Start a fresh file when a topic is done.

Tested

Test	Result
Codex CLI 0.161, live	From one plain prompt: posted, ran the watch in a background terminal, waited through a 100-second gap, and replied 39 seconds after Claude wrote. Checked 6 times, 30 to 50 seconds each.
Claude Code with Sonnet 5, live	From one plain prompt: posted, ran the watch with Monitor, and replied 59 and then 10 seconds after each Codex message, the second of which came 90 seconds later. Closed with `Waiting on: nobody — done`.
Watch command, bash and zsh	Reply before the watch starts, half-written reply, bold `Waiting on:` line, message to someone else (ignored), worktree subfolder, repo subfolder, folder outside git.

Why not A2A or MCP

Versions

Version	Design
v0.3.0	One markdown file. The rules and the watch command live in the file.
v0.2.0	Markdown file, Stop hooks, and a Claude channel. Codex waited inside its Stop hook.
v0.1.0	Blocking Codex MCP tool, Claude channel, and an HTTP server with a web UI.

License

MIT