DEV Community: Randy Northrup

Where Muse Spark Code is going: bring your own models, agent teams and every editor

Randy Northrup — Wed, 07 Oct 2026 00:26:04 +0000

Muse Spark Code is a free, open source coding agent for Meta's Muse Spark. I've been building it in public, and a lot of people have asked the same question: where is it going? This is the answer, along with a quick tour of what it already does.

What it is today

One agent, every editor. Muse Spark Code runs as an ACP agent in Zed, JetBrains IDEs, Neovim and Emacs, and as an extension for Cursor, Windsurf, VSCodium and VS Code. It works with your Muse subscription through the Muse Code CLI, or with a Meta Model API key.

A few things make it different from a plain chat panel:

Subagents you can watch. When work is delegated, every subagent shows up on a live agent map with its role, status, token use and its own transcript, so you can follow and steer it.
It checks its own web changes. The agent can open your local dev server in an isolated headless browser, click and type, and read console errors and failed requests before it tells you it's done.
Hardware, too. A bundled Muse Gadgets skill builds, flashes and monitors ESP32 boards.
Questions don't block you. When the agent needs your input, the question waits in a card while you keep working. You can answer later, or explain what you want instead.
No telemetry. It's MIT licensed, has no hosted server of its own, ships in 15 languages and is checked against WCAG 2.2 AA.

Where it's going

These are roughly in order. There are no dates; each one ships when it's ready.

Save and share prompts and chats. A personal prompt library you can use in any project, with right-click Save on your own messages and prompts with variables. Chats export as Markdown, HTML or JSON, with an exact preview and redactions highlighted.
Bring your own models. OpenAI, Anthropic and Google Gemini keys, local models, and ChatGPT and GitHub Copilot sign-in, alongside Muse Spark. Each feature turns on only for models that support it.
Agent teams. Research, design, engineering, QA, code review, docs and marketing roles, each with its own caps, orchestrated by the main agent.
A usage and cost page in every editor. One local journal of what each model and feature cost.
/legal. A read-only scan for licensing and copyright issues. Fixes happen only after you choose them.
Every editor, every feature. JetBrains IDEs, Visual Studio, Eclipse, Xcode, Zed, Neovim, Emacs, Sublime Text and Qt Creator get the full feature set.
Makers. Arduino, ESP32 and Raspberry Pi boards, 3D printers and CNC, with simulators. Every physical action needs your approval at the OS level, and Stop always works.
Muse Judge. A second-opinion judge that can check any agent's work.
/report. Deterministic, redacted project reports you can schedule, save and compare over time.
Scheduled prompts. Unattended runs on a schedule or a trigger, with a choice of how results are delivered.
A resource governor. CPU and memory limits, so agents never bog down your machine.
Video, audio, documents and screen recordings as input, with transcription where the model supports it.
Several accounts per provider, with usage thresholds.
A credential vault. Agents use secrets through a broker and never see them.
A capacity estimator. How long a plan will take on your machines, and the cheapest or fastest setup to run it.
A design-language refresh across the whole UI.

How it gets built

Most of the implementation is done by a fleet of coding agents with a lot of structure around them. Claude plans each milestone and splits it into small lanes, Codex and Muse agents build those lanes across several machines, and a different agent reviews every lane before it can merge. That's how a solo developer can ship several releases in a day without losing track of quality.

Try it

GitHub: https://github.com/RandyNorthrup/muse-spark-code
Open VSX (Cursor, Windsurf, VSCodium): https://open-vsx.org/extension/RandyNorthrup/muse-spark-code
VS Code Marketplace: https://marketplace.visualstudio.com/items?itemName=RandyNorthrup.muse-spark-code
ACP agent for other editors (npm): muse-spark-code-acp

If Muse Spark Code is useful to you, a star on GitHub helps other people find it. Issues and feature requests are welcome too, especially if one of the roadmap items matters to you.

Muse Spark Code is unofficial and not affiliated with or endorsed by Meta. "Muse Spark" and "Muse Code" are Meta trademarks.

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

Randy Northrup — Mon, 05 Oct 2026 18:37:22 +0000

Meta's Muse Code CLI is a capable coding agent, but it lives in the terminal. I wanted it in my editor, with the same kind of experience I get from other agents, so I built Muse Spark Code: an unofficial, MIT-licensed coding agent for Meta's Muse Spark.

It just reached version 0.13.0 and is closing in on 1,000 downloads, so this seemed like a good time to explain how it works and how it gets built.

One agent, many editors

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.
An editor extension for Cursor, Windsurf, VSCodium, VS Code and other editors built on the same platform, published on both the Marketplace and Open VSX.

Either way you get the same agent: a chat that streams answers, reads and edits files with reviewable diffs, runs commands behind permission modes, and delegates to subagents.

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.

What's new in 0.13.0

Browser checks on your local dev server

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. On the Model API backend it also receives the page's screenshot. The browser runtime is verified and downloaded once, on your request.

This closes a loop that coding agents usually leave open: the agent can now see that its change broke the page instead of assuming it worked.

A Muse Gadgets skill for hardware

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. A command that times out still returns its captured output and leaves no process behind on Windows, macOS and Linux.

Secret handling

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. Diagnostic notices redact known secret shapes too.

Subagents you can watch

When Muse Code delegates work, each subagent shows up as a row, and an agent map lists every agent with its role, status, token use and its own transcript. You can follow and steer them while they run instead of waiting for a final answer.

Built by a fleet of agents

The part people ask about most is how a solo developer ships several releases a day. The answer is that most of the implementation is done by agents, with a lot of structure around them:

Claude orchestrates. It plans each milestone and splits it into small, independent lanes.
Codex and Muse agents build. Lanes are dispatched across several machines: my PC, a Kubuntu VM, a Mac mini and a Windows VM with a GPU.
Every lane is checked every 30 minutes: is it alive, committing, and on its ETA? A lane that crashes restarts from its worktree and keeps its uncommitted work.
A different agent reviews every finished lane before it can merge. A failed review starts a fix round immediately, and a third failed round means the part is redesigned rather than patched again.
Work lands in batches: one merge, one full quality check, one pull request and one CI run per release.

The lesson for me has been that the coding agents are not the hard part. The scheduling and review discipline around them is what makes the output trustworthy.

The rest

No telemetry and no hosted server of its own
15 languages
Checked against WCAG 2.2 AA in every default theme
Free and MIT licensed

Try it

The source, docs and full release notes are on GitHub:

https://github.com/RandyNorthrup/muse-spark-code

If you use an ACP client other than Zed, I'd especially like to hear how it works for you. Issues, pull requests and stars are all welcome.

Muse Spark Code is unofficial and not affiliated with or endorsed by Meta. "Muse Spark" and "Muse Code" are Meta trademarks.