codex-collab

codex-collab

English | 中文

Collaborate with Codex from Claude Code. Run tasks, get code reviews, do parallel research, all without leaving your Claude session.

codex-collab is a Claude Code skill that drives Codex through its app server JSON-RPC protocol. It manages threads, streams structured events, handles tool-call approvals, and lets you resume conversations.

Why

• Native peer messaging — The workspace broker registers Codex as a peer in Claude Code's cross-session messaging, so Claude sessions message Codex directly (SendMessage) and Codex answers back — including asking Claude questions mid-task via a collab.consult tool. No CLI in the hot path.
• Structured communication — Talks to Codex via JSON-RPC over stdio. Every event is typed and parseable.
• Event-driven progress — Streams progress lines as Codex works, so Claude sees what's happening in real time.
• Review automation — One command to run code reviews for PRs, uncommitted changes, or specific commits in a read-only sandbox.
• Thread reuse — Resume existing threads to send follow-up prompts, build on previous responses, or steer the work in a new direction.
• Approval control — Configurable approval policies for tool calls: auto-approve, interactive, deny, or Codex's Guardian auto-reviewer (--approval auto).
• Two-way ask channel — Codex can ask a question mid-turn (ask) and keep working once the answer arrives (answer); next blocks until something needs attention. Fail-open: an unanswered question never stalls a run.
• Live observability — run --detach hands a long task to a detached runner; follow --watch is a purpose-built live view that tracks every run in a terminal pane.
• Memory isolation — Threads created by codex-collab are excluded from Codex's memory feature by default, so agent-driven sessions don't shape Codex's learned picture of how you work. Opt back in with --memory (see Options for details).

Installation

Requires Bun >= 1.0 and Codex CLI (npm install -g @openai/codex) on your PATH. Tested on Linux (Ubuntu 22.04), macOS, and Windows 10.

git clone https://github.com/Kevin7Qi/codex-collab.git
cd codex-collab

Linux / macOS

./install.sh

Windows

powershell -ExecutionPolicy Bypass -File install.ps1

After installation, run codex-collab health to verify. If the command is not found, ~/.local/bin is not on your PATH yet: reopen your terminal, add the export line the installer prints, or use the full path ~/.local/bin/codex-collab health.

Tip You can hand the install to an agent. Running it outside Claude Code, for example from Codex, means the skill is ready the next time you start Claude. It will ask permission to write outside the repository.

Where things get installed

The installer builds a self-contained bundle plus a binary shim. On Linux and macOS these go to ~/.claude/skills/codex-collab/ and ~/.local/bin. On Windows they go to %USERPROFILE%\.claude\skills\codex-collab\, and install.ps1 adds the shim to your PATH.

Claude discovers the skill automatically, including in a session that is already running. The exception is your first skill: if ~/.claude/skills/ did not exist before, restart Claude Code once so the new directory is watched.

Upgrading

An installed codex-collab can update itself — no manual git pull needed:

codex-collab update            # show the latest release and changelog, confirm, then install
codex-collab update --check    # report only, install nothing

update fetches the latest release, rebuilds it locally, and reinstalls. Nothing is installed without your confirmation: an interactive y/N prompt, or an explicit --yes when there is no terminal to ask. run, review, and health print a one-line notice when a newer release exists, but never install anything themselves.

Upgrading manually still works, and is the only way to update a dev install (install.sh --dev):

git pull
./install.sh    # Windows: powershell -ExecutionPolicy Bypass -File install.ps1
codex-collab health

More on upgrading

update --skip mutes notices for one release, and CODEX_COLLAB_NO_UPDATE_CHECK=1 turns the release check off entirely. The local skill-drift check is offline and stays on regardless.

If the installed SKILL.md falls out of step with the binary or your template set, codex-collab skill sync shows the pending diff and applies it once you confirm.

Both upgrade paths replace the skill bundle and the binary shim. Everything under ~/.codex-collab/ is preserved: configuration, templates, thread history, and run logs. Treat ~/.claude/skills/codex-collab/ as installer-managed, since manual edits there can be overwritten on upgrade.

Development mode

Use --dev to symlink source files for live-reloading instead of building a bundle:

# Linux / macOS
./install.sh --dev

# Windows (may require Developer Mode or an elevated terminal for symlinks)
powershell -ExecutionPolicy Bypass -File install.ps1 -Dev

Quick Start

# Run a prompted task
codex-collab run "what does this project do?" -s read-only --content-only

# Code review
codex-collab review --content-only

# Resume a thread
codex-collab run --resume <id> "now check error handling" --content-only

# Long task: detach it, watch it live in another pane
codex-collab run "large refactor" --detach --approval auto
codex-collab follow --watch

CLI Commands

Command	Description
run "prompt" [opts]	Start a thread, send a prompt, wait, print the output (run - reads the prompt from stdin)
review [opts]	Code review (PR, uncommitted, or a specific commit)
threads [--json] [--all]	List threads (--discover scans the server and marks threads open or active on the app-server, --session limits to this session)
follow [id]	Live view of a running thread in your own terminal pane. Without an ID it attaches to the active run; --watch keeps following each new run
output <id> [--last]	Full log for a thread (--last: only the latest turn's output)
kill <id> [--clear]	Stop a running thread. An active goal is paused first; --clear abandons it
peer [up]	Show the native-messaging peer's status; peer up starts the broker (and with it the peer)

The commands Codex uses to reach Claude (peers, send, task) are under Peer messaging.

Questions and approvals

Command	Description
ask "question"	Invoked by Codex mid-turn to ask a question and wait for the answer. --timeout <sec> sets the deadline (default 600). Fails open: on expiry it tells Codex to proceed on its own judgment and exits 0
answer <id> "text"	Answer a pending question (answer <id> - reads from stdin)
questions [id]	List pending questions in this workspace; with an ID, show the full text
next	Block until something needs attention, print it in full with how to respond, then exit
approve <id>	Approve a pending request
decline <id>	Decline a pending request

Inspection and configuration

Command	Description
progress <id>	Recent activity (tail of the log)
peek <id>	Recent conversation slice from the server
config [key] [value]	Show or set persistent defaults
models	List available models
templates	List available prompt templates

Maintenance

Command	Description
delete <id> [--purge]	Archive a thread (recoverable via codex unarchive) and delete local files; --purge deletes it server-side instead
clean	Delete old logs and stale mappings
skill sync [--yes]	Regenerate the installed SKILL.md when it drifts from the binary or template set. Prints the diff, applies only on confirmation
update	Check for a newer release and install it with confirmation. See Upgrading
health	Check dependencies and authentication, name the app-server the broker runs on, and report Codex-side skill and exec-policy status
version	Print version (also -v/--version before a command)

Options

General

Flag	Description
-d, --dir <path>	Working directory
-m, --model <model>	Model name (default: auto — latest available)
-r, --reasoning <level>	none, minimal, low, medium, high, xhigh, max, ultra (default: auto — highest the model supports, up to xhigh)
-s, --sandbox <mode>	read-only, workspace-write, danger-full-access (default: workspace-write). review rejects this flag: reviews always run read-only
--resume <id>	Resume existing thread
--approval <policy>	Approval policy: never, on-request, on-failure, untrusted, auto (default: never). auto: Codex's Guardian reviewer approves or denies each request autonomously — never blocks on a human; decisions stream as Guardian lines. review rejects this flag: Codex locks review sub-agents to never, so it could never take effect
--memory	Let Codex's memory feature learn from threads this run creates. Default: created threads get thread/memoryMode/set mode=disabled; resumed threads are never touched (the flag is persistent per-thread, and a thread you created yourself should keep feeding your memory). Governs Codex's local memory consolidation (~/.codex/memories) only — the personality feature is explicit user config (not learned) and unaffected. Persistent form: config memory true
--timeout <sec>	Turn timeout (default: 3600, max 2147483). When a goal is active it scopes the whole goal, and expiry pauses the goal before exiting. For ask: answer deadline (default: 600); for next: wait deadline (default: wait indefinitely)
--	End of options; remaining arguments are treated as prompt text

run

Flag	Description
--detach	Return once the turn is running; watch with follow <id>. Turn lifetime is decoupled from the invoking shell
--template <name>	Prompt template (user ~/.codex-collab/templates/ or built-in)
--goal <objective>	Create the thread's goal before the first turn (replaces the objective on --resume); requires goals = true in ~/.codex/config.toml. A prompt is still required — it is turn one, while the goal is the standing objective. With --template collab the objective also gets a one-line ask-channel note — re-injected into every continuation turn, so channel awareness survives long goals. review rejects this flag: a review is a single turn on an ephemeral thread
--budget <tokens>	Token budget for --goal. Size generously — usage counts each turn's full context, so a single small turn can consume ~60k. review rejects this flag
-	Read the prompt from stdin

review

Flag	Description