Muse Spark Code: Meta's Muse Spark as a coding agent in any ACP editor

Description and use

Muse Spark Code extends Muse Code, enabling it as an agent in editors with browser debugging, ESP32 hardware support, and secure secret handling.

Problem

Meta’s Muse Code CLI is terminal‑only, limiting editor integration and user experience.

Outcome

Muse Spark Code extends Muse Code into editors, adding browser debugging, hardware support, and secret handling, achieving near‑real‑time coding agent integration without telemetry.

Workflow

• 1. Install Muse Spark Code via npm or as a VS Code extension.
• 2. Configure with a Muse subscription or Meta Model API key (key never exposed to CLI).
• 3. Use the editor’s ACP client or extension to launch the agent.
• 4. Interact through chat to request code changes, run commands, or debug.
• 5. Agent streams responses, edits files, and shows reviewable diffs.
• 6. For debugging, agent can open local dev server pages in a headless browser and report console errors.
• 7. For hardware, agent can build, flash, and monitor ESP32 devices via Muse Gadgets skill.

Evidence

Meta's Muse Code CLI is a capable coding agent, but it lives in the terminal.

Source

I wanted it in my editor, with the same kind of experience I get from other agents, so I built Muse Spark Code

Source

Muse Spark Code ships two ways: An ACP agent. The Agent Client Protocol lets an editor drive any compatible agent, so Zed, JetBrains IDEs, Neovim, Emacs and other ACP clients can run Muse Spark Code directly.

Source

It runs on your Muse subscription through the Muse Code CLI, or on a Meta Model API key. The API key is never handed to the CLI.

Source

After a web change, the agent can open a page on your local dev server in an isolated headless browser, click and type, and read console errors and failed network requests.

Source

Meta recently released Muse Gadgets, an SDK for building ESP32 devices that work with Muse. Muse Spark Code now bundles a skill that walks the model through ESP32 build, flash and bounded serial monitoring.

Source

Detected tokens are redacted, a prompt that contains a secret pauses for your decision, and a shell approval that would expose one only offers one-time consent.

Source