# Muse Code (slim)

> Muse Code is Meta's coding agent for the terminal and CI, built for its Muse models. It plans, edits files and runs commands under approvals and an OS sandbox, interactively, headless through muse exec or from an SDK.

- Full: https://www.anchorterminal.com/tools/muse-code.md (~6,950 tokens) · this version ~1,580 tokens · JSON https://www.anchorterminal.com/tools/muse-code.json · canonical https://www.anchorterminal.com/tools/muse-code
- Index: https://www.anchorterminal.com/llms.txt · API: https://www.anchorterminal.com/api/v1/index.json · Updated: 2026-10-12

**C · 61.3/100 · rank #487 of 993 · #10 in Agent harnesses · not agent-ready · confidence medium**


Assessment: Muse Code runs with approvals and an OS sandbox on by default, per-host network approval, an append-only event log and a headless mode with documented exit codes. It is the only route to Muse Wonder Full. It is closed source with undated release notes and no free route, and subscriptions allow training on code unless Contributor mode is turned off.

## Facts

- Kind: Agent harness · vendor: Meta · category: Agent harnesses · legal entity: Meta Platforms, Inc. · provenance 68/100
- Packages: npm `@muse-code/sdk`, pypi `muse-code-sdk`
- Auth: OAuth or key · pricing: Paid · x402: no · licence: Proprietary binary, used under the Meta Model API Terms of Service, which cover Meta's coding harness. Subscriptions are also under the Meta Subscription Terms of Service. The TypeScript and Python SDKs in the linked repository are MIT.
- Probe metrics: not measured yet (probes haven't run)
- Models: Meta's Muse models through Meta Model API. The default is `muse-spark-1.2`. Muse Wonder (`muse-wonder-1.0`) on pay-as-you-go and every plan, Muse Wonder Full on the Power Usage and Maximum Usage plans only. No other provider is documented.
- Install: One-line installers for macOS and Linux (`install.sh`) and Windows (`install.ps1`) that fetch a native binary from a release channel and check SHA-256 sums. The product page says macOS and Windows, and the docs add Linux.
- Permission profiles: Ask me, Auto-review (the default, an automated judge that falls back to asking) and Unrestricted (`--yolo`). Compound shell commands are reviewed stage by stage. Managed configuration can set policy.
- Sandbox: On by default. Seatbelt on macOS, a bundled bubblewrap on Linux and a Windows sandbox that may ask for administrator approval. Network modes proxy-only (default), restricted and unrestricted.
- MCP client: stdio and streamable HTTP servers in `mcp_servers`, OAuth 2.1 sign-in with `muse mcp login`, and named toolsets. Read-only tools skip the approval prompt. MCP tools are not sandboxed.
- Headless: `muse exec` with `--json` JSONL events, `--prompt-file`, `--max-model-steps`, `--session-id` to continue a run and `muse export` for a session record. Subagents in isolated git worktrees and workflows for parallel work.
- Plans: Everyday Usage $5, High Usage $15, Power Usage $50 and Maximum Usage $200 a month, with usage set as multiples of Everyday (10 to 50 prompts every 5 hours). Contributor mode is on by default for subscribers.
- Release: 1.4.6, the version the muse-stable channel served on 12 October 2026. The SDKs are released in step, with @muse-code/sdk 1.4.6 published on 11 October 2026. The CLI release notes have no dates.
- Prices: Everyday Usage plan $5 per month (plan); High Usage plan $15 per month (plan); Power Usage plan $50 per month (plan); Maximum Usage plan $200 per month (plan)

- Scores: Reliability 45, Performance pending, Schema & documentation 84, Agent ergonomics 79, Security & auth 67, Payments & pricing 20, Task success pending, Maintenance & community 75, Transparency & trust 57 · total over the 7 assessed categories
- Why: Reliability, Local-package reading. · Schema & documentation, Framework reading. · Agent ergonomics, Framework reading, adapted to a harness driven by a pipeline. · Security & auth, Framework reading (telemetry defaults, approvals, guardrails, sandboxing), five lines. · Payments & pricing, Harness reading of the published rubric. · Maintenance & community, 1.4.6, served by the muse-stable channel on 12 October 2026 and published as SDK 1.4.6 on 11 October (30). · Transparency & trust, Closed-source binary under the Meta Model API Terms of Service, with MIT SDKs (15).
- Sources: 19, open questions: 6, both in the full twin
- Capabilities: agent.harness, agent.mcp-client, agent.multi-agent
- JSON: https://www.anchorterminal.com/api/v1/tools/muse-code.json
- Verify (for the vendor): the badge `https://www.anchorterminal.com/badges/muse-code.svg` or a link to https://www.anchorterminal.com/tools/muse-code from a page on meta.ai or one of its subdomains, or the README of github.com/meta-models/muse-code-sdk, then `POST https://www.anchorterminal.com/api/v1/verify` `{"slug", "url"}` or `verify_listing` at /mcp; re-checked weekly, no effect on the grade. Snippets in the full twin.

## Before you call it

1. Set `META_API_KEY` for headless runs. Browser sign-in isn't available in CI or for Meta Managed Accounts
2. Gate on your own test command, not the exit code. `muse exec` returns 0 when the turn completes even if tests still fail
3. Pass `--max-model-steps` on unattended runs so a stuck task can't loop
4. Prefer `--disable-approval` to `--yolo` in CI so the sandbox stays on. On Linux the runner needs a working bubblewrap and a non-musl build
5. Turn Contributor mode off in `/settings` before a subscription session touches private code

## Connect

```bash
curl -fsSL https://dev.meta.ai/install.sh | sh   # Windows: irm https://dev.meta.ai/install.ps1 | iex
```

## Similar tools

| Tool | Grade | Score | Shared capabilities | Slim |
| --- | --- | --- | --- | --- |
| Kilo Code CLI | BB | 75.4 | agent.harness, agent.mcp-client, agent.multi-agent | https://www.anchorterminal.com/tools/kilo-code-cli.min.md |
| goose | BB | 73.9 | agent.harness, agent.mcp-client, agent.multi-agent | https://www.anchorterminal.com/tools/goose.min.md |
| Qwen Code | BB | 72.4 | agent.harness, agent.mcp-client, agent.multi-agent | https://www.anchorterminal.com/tools/qwen-code.min.md |
| OpenHands | BB | 70.8 | agent.harness, agent.mcp-client, agent.multi-agent | https://www.anchorterminal.com/tools/openhands.min.md |
| OpenCode | B | 67.7 | agent.harness, agent.mcp-client, agent.multi-agent | https://www.anchorterminal.com/tools/opencode.min.md |


## Panel reviews (0, desk reviews from public material, no calls made)