Muse Code

Connect AIWG to Muse Code (experimental, trust-workspace flow)

Connect AIWG to Muse Code

Prompt-first procedure: Describe the outcome you want in your agent conversation. The agent should select and load the appropriate AIWG assets, explain material changes, request any needed approval, and report verification evidence. Exact commands and flags appear only in the CLI reference.

Status: Stable.

For the complete first-time journey, start with Install, Connect, and Verify.

Muse Code is the provider in this guide — Meta's terminal/CI coding agent built on Muse Spark. Complete the safe Node.js setup first if `node` or `npm` is unavailable.

1. Install Muse Code

"Check whether Muse Code is installed on this machine and ready for AIWG."

Your assistant confirms the `muse` binary is on your PATH.

Install Muse Code for your platform and confirm the `muse` binary is on your PATH:

Command reference

muse the version option

2. Install AIWG and deploy

"Install AIWG in this project and deploy it for Muse Code, then show me the verification result."

Your assistant installs AIWG and runs the agent-owned use operation, which refreshes AIWG's shared project context and prints a verification result.

From a terminal opened in the project's main folder, install AIWG and deploy the complete system:

Command reference

Use AIWG to complete this documented outcome: Command reference
Have it inspect the current state, explain the plan, ask before material
changes, and report the result with verification evidence.

The deployment command refreshes AIWG's shared project context and prints a verification result.

3. Trust the workspace, then start a new session

"I trusted this workspace in Muse. Load the deployed bridge and skills and tell me what you can see."

Your assistant starts a new session so Muse loads the deployed bridge and skills from the trusted workspace.

"I trusted the workspace but the AIWG skills are still missing — what now?"

Saying the skills are still missing points to the fix: start a new Muse session or re-read skills, since reloading an IDE window does not apply to this provider.

Muse Code loads project `AGENTS.md` only after the workspace is explicitly trusted. When Muse shows the first-run trust prompt, trust the workspace, then start a new Muse session (or re-read skills) so the deployed bridge and skills load. Reloading an IDE window does not apply here — this provider is not Cursor.

4. Work discover-first

"What AIWG skills and rules are available for this task before you start?"

Your assistant runs the agent’s capability search first and reads the deployed skill or rule before improvising.

Muse skills deploy natively to `<project>/.agents/skills/`; agents, commands, and rules stay indexed. Before improvising, run:

Command reference

Use AIWG to complete this documented outcome: Command reference
Have it inspect the current state, explain the plan, ask before material
changes, and report the result with verification evidence.

Then ask Muse Code to verify AIWG by reporting the project root, provider files it can read, installed frameworks, and one useful next action.

Try one small task immediately after verification:

"Review this project's README and getting-started docs for unclear positioning, missing setup steps, or unsupported claims. Save the three highest-priority fixes with file references and a recommended next edit at .aiwg/marketing/brand/audit/readme-review.md. Leave the reviewed files unchanged."

Your assistant follows the AIWG bootstrap into `WORKSPACE.md` and `AIWG.md` or the provider-specific adapter and produces a concrete review you can inspect.

Success means Muse Code names the intended project, follows the AIWG bootstrap into `WORKSPACE.md` and `AIWG.md` or the provider-specific adapter, and produces a concrete review you can inspect.

User-scope skills

"Deploy AIWG skills for Muse Code across all my projects, not just this one."

Your assistant deploys user-level skills, resolved at deploy time, without touching other provider paths.

For user-level skills across projects:

Command reference

Use AIWG to complete this documented outcome: Command reference
Have it inspect the current state, explain the plan, ask before material
changes, and report the result with verification evidence.

Skills land in `$XDG_CONFIG_HOME/muse/skills` (default `~/.config/muse/skills`), resolved at deploy time. AIWG never writes to `~/.muse`, foreign provider paths, or `~/.agents/skills`.

Sessions

"Share this Muse session's history with the operator."

Your assistant shares an explicit exported JSON document rather than assuming a native log location.

Session history is export-first: share an explicit `muse export` / `/export trajectory` JSON document with the operator rather than assuming a native log root. AIWG never scrapes `~/.muse` or similar homes for sessions.

For advanced flags, compatibility notes, and recovery details, see the Muse Code operational reference.