OpenAI Codex

OpenAI Codex can use Mirage through either an installable TypeScript plugin or a FUSE mount. The Python integration FUSE-mounts a workspace as a normal host directory, so Codex’s built-in filesystem and shell tools operate on Mirage without a separate tool server.

Install

uv add 'mirage-ai[fuse]'

Then install OpenAI Codex separately.

Usage

from mirage import Mount, MountBackend, MountMode, Workspace
from mirage.vfs.ram import RAMVFS

with Workspace(
    {"/": Mount(RAMVFS(), mode=MountMode.WRITE, backend=MountBackend.FUSE)}) as ws:
    print(f"cd {ws.fuse_mountpoint} && codex")
    input("Press Enter when done...")

The mountpoint behaves like a regular directory. Codex’s built-in file and shell tools all dispatch through Mirage’s ops layer. For the Codex app, open the mountpoint as the project folder instead of starting codex from it.

FUSE or plugin?

Use Python FUSE when you want Codex’s built-in filesystem tools to see Mirage as an ordinary directory. This works in the CLI and app, but requires an OS FUSE driver and does not add Mirage’s agent-level stale-write check. Use the TypeScript plugin when you want named Mirage tools, Pi-style stale-write protection, and plugin installation through Codex. The plugin uses Mirage’s standard MCP adapter internally.

Getting Started

AI Agents

VFS

CLIs

Watch

Runtimes

Install

Usage

FUSE or plugin?

​Install

​Usage

​FUSE or plugin?

Install

Usage

FUSE or plugin?