Your assistant. Your development environment.

Your remote workspace.
Right here in the conversation.

Let your assistant read project files, run code, and work with Codex on your host—without copying terminal output back and forth.

Understand the project

Inspect files, read contributor guidance, and find saved Codex sessions.

Get work done

Run Python or shell commands, or delegate a task to a remote Codex agent.

Keep what you learn

Build reusable session knowledge. Browse it in OpenWiki or search it through your assistant.

A client agent’s first impression

“The user could point me at a remote project, and I could inspect the source material myself.”

In one RemoteCodex MCP smoke test, Codex discovered the host, ran 6 * 7, and read a project’s AGENTS.md.

Codex · Excerpt from a client-agent account. Model-driven tasks and interrupted-connection recovery were not tested in that session. Not an official OpenAI endorsement.

Use with Codex

Set MCP_TOKEN in the environment that launches your local Codex client, then add your server:

codex mcp add remotecodex \
  --url https://YOUR_HOST/mcp \
  --bearer-token-env-var MCP_TOKEN