OpenAI Codex

Way	Codex gets	Needs
CLI	Mirage’s seven tools, beside its own	the `mirage` CLI
SSH	its own tools, run in the workspace	the daemon’s SSH server
FUSE	its own tools, on the workspace as a folder	a FUSE driver

Install

uv tool install 'mirage-ai[fuse]'

Install Codex separately.

Workspace

workspace.yaml

mode: write
mounts:
  /:
    vfs: ram
    backend: fuse                 # FUSE only
    mountpoint: /tmp/mirage-demo  # FUSE only
  /s3:
    vfs: s3
    mode: read
    config:
      bucket: my-bucket

mirage workspace create workspace.yaml --id demo

CLI

codex mcp add mirage -- mirage mcp -w demo

Codex gets shell, read, write, edit, ls, grep and glob from the workspace’s MCP server. Or install the Mirage plugin, which adds a skill and reads .mirage/workspace.yaml from your project instead:

codex plugin marketplace add strukto-ai/mirage
codex plugin add mirage@mirage

SSH

Turn on the daemon’s SSH server, then add it as a Codex environment:

~/.codex/environments.toml

default = "mirage"
include_local = false

[[environments]]
id = "mirage"
program = "ssh"
args = ["-T", "-p", "2222", "demo@127.0.0.1", "-s", "codex-exec"]

codex -C /

Codex’s shell commands, file reads and patches run in the workspace, under the SSH key’s profile. This works in the Codex CLI, not the app. The protocol is experimental and follows Codex 0.147.

FUSE

cd /tmp/mirage-demo && codex

Every file and shell call Codex makes goes through Mirage: /tmp/mirage-demo/s3 is the bucket. In the Codex app, open the mountpoint as the project folder. mirage workspace delete demo unmounts it. On macOS the daemon holds one FUSE mount at a time.

Getting Started

Access

AI Agents

VFS

CLIs

Watch

Runtimes

Install

Workspace

CLI

SSH

FUSE

​Install

​Workspace

​CLI

​SSH

​FUSE

Install

Workspace

CLI

SSH

FUSE