Skip to content
aicoolies logo
Meta logo

Muse Code

Terminal AI coding agent powered by Meta Muse Spark models

Muse Code is Meta autonomous terminal coding agent powered by the Muse Spark model family. Designed for multi-agentic software engineering, it inspects repositories, plans architectural changes with approval gating, executes edits across multiple files, and maintains crash-resilient session event logs.

About Muse Code

Muse Code is Meta autonomous command-line coding agent designed to execute end-to-end software engineering workflows directly from the terminal. Built for multi-agentic development, the tool moves beyond simple inline completions by inspecting entire repository architectures, formulating explicit implementation strategies with user approval checkpoints, modifying source files across multiple packages, and verifying modifications via local test runners and linters.

The underlying intelligence comes from Meta's Muse Spark model family, which has a 1-million-token context window and accepts text, image, video, audio and PDF input; Muse Spark 1.3 is the latest hosted version, and Muse Code's default model is muse-spark-1.2, switchable with --model or /models. Meta also publishes open-weight models: Muse Glimmer, a 30B multimodal model distilled from Muse Spark for local and edge deployment under Apache 2.0, and Muse Spark Open (see below). The agent utilizes OS-level sandboxing and explicit workspace trust boundaries to protect local developer environments from unintended execution side effects.

Muse Spark Open and on-device use: on October 15, 2026 Meta released the weights of Muse Spark Open, an open-weight version of Muse Spark with 816B total parameters (42B active), a 1M-token context window and text, image, audio and video input, under the Muse Model License with limited commercial terms. Meta reports testing agentic workflows running entirely on-device using Muse Code powered by Muse Spark Open, including shipping iOS changes to the open-source Ice Cubes app and driving DaVinci Resolve. Local use needs serious hardware: Meta's run-locally recipe calls for two Mac Studios with M5 Ultra and 512 GB of unified memory each (or four with 256 GB), macOS 26.2 or later, MLX 0.30.1 or later and direct Thunderbolt 5 links, serving the roughly 537 GB NVFP4 checkpoint through mlx-lm as an OpenAI-compatible API. The full BF16 checkpoint is about 1.65 TB, and Meta's GPU serving recipes target 8x B300 (BF16) or 4x B300 (NVFP4). As of October 16, 2026 the Muse Code docs and changelog do not describe a dedicated local mode for connecting Muse Code to a self-hosted Muse Spark Open server; a standard install signs in to Meta Model API with a browser sign-in or API key. Muse Spark Open is also offered by hosted providers including Together AI, Fireworks AI, Baseten, Modal, Groq, Nebius and OpenRouter.

Muse Code features a multi-agent orchestration architecture that coordinates persistent background subagents to perform parallel tasks such as dependency analysis, test execution, and documentation lookup without blocking interactive developer flow. Built-in terminal commands include /plan for generating approval-gated change strategies, /grill for stress-testing and probing architectural plans, and /goal for driving complex objectives to completion. In addition, multi-session setups can exchange status and artifacts over Unix domain sockets.

Session reliability and transparency are maintained through an append-only local event log that records every model invocation, tool execution, and source patch, providing crash safety and replay-exact auditability. Muse Code operates on token-based pass-through billing via the Meta Model API on dev.meta.ai, offering both private Standard Tier and training-opted Contributor Tier access with zero upfront license fees.

Pricing & Platform Specs

Pricing Summary

Muse Code is billed either pay-as-you-go through a Meta Model API key or with a flat-rate Muse Code subscription. Pay-as-you-go Muse Spark rates per 1M tokens: Standard mode $1.25 input, $0.15 cached input, $4.25 output (content not used for training); Contributor mode $0.10 input, $0.002 cached input, $0.20 output in exchange for letting Meta train on your content. Muse Wonder (Standard only) costs $4.00 input, $0.20 cached input and $20.00 output per 1M tokens. Web search grounding is $2.50 per 1,000 queries. Subscriptions come in four tiers (Everyday, High, Power and Maximum Usage); prices are shown during onboarding and are not published in the docs, and subscriptions default to Contributor mode, which can be turned off.

full pricing breakdown →

Supported Platforms

macOS, Linux, Windows (native; PowerShell)

Explore categories, tags & use cases
Claude Code logo

Claude Code

Top Pick

Anthropic's agentic coding CLI

Anthropic's agentic CLI coding tool that delegates complex tasks to Claude directly from the terminal. Understands entire codebases via automatic context gathering, edits multiple files, runs shell commands, and manages Git workflows autonomously. Supports CLAUDE.md for persistent project instructions, integrates with VS Code and JetBrains, and runs on current Claude models such as Opus 5.5 and Sonnet 5.5 for complex architecture work. Built for terminal-first developers.

freemium
Codex logo

Codex

Top Pick

OpenAI coding agent for app, editor, terminal, and cloud work

Codex is OpenAI's coding agent for software development across the Codex app, editor, terminal, and cloud tasks. It helps write, review, debug, refactor, and automate code, with ChatGPT plan access for managed surfaces and API-key usage for CLI, SDK, and IDE workflows. The open-source CLI and SDK support local repository work, while cloud features add GitHub review, Slack/Linear integrations, worktrees, skills, MCP, and automations.

paidTelemetry

Enterprise-grade AI coding agent system by Factory

Factory's AI coding agent, available in the Factory App (desktop, web and mobile), the Droid CLI, headless droid exec and Factory-hosted Droid Computers. Custom droids (subagents) with their own prompt, model and tool policy take focused tasks, and Missions orchestrate large multi-feature projects. Works with Anthropic, OpenAI and other native models, plus BYOK, open-source and local models. Enterprise-focused, with automated code review, QA, docs and incident response.

freemium
Devin logo

Devin

Top Pick

Autonomous AI software engineer by Cognition

Devin is Cognition's managed AI software engineer for delegating engineering tasks to cloud and desktop agents. It can plan work, navigate codebases, write and run code, test changes, open PRs, review/autofix issues, and collaborate through GitHub, GitLab, Bitbucket, Linear, Jira, Slack, and Teams. Current Devin surfaces include Devin Cloud, Devin Desktop, Devin CLI, Devin Review, Windows VM support, DeepWiki, Ask Devin, and team/enterprise controls.

freemium

The AI-first code editor

Cursor is an AI coding agent for building software, available as a desktop editor, a CLI and cloud agents that run in isolated VMs. Its Agent edits files, searches the codebase and runs terminal commands in Agent, Plan and Ask modes; Premium and higher tiers add Automations, Bugbot pull-request reviews and Origin code hosting. Sold within The X Subscription, it offers Cursor's own models (Grok and Composer), third-party models on Premium and above, and Privacy Mode on every plan.

freemiumTelemetry

Community experience

FAQ

What is Muse Code?

Muse Code is Meta autonomous terminal coding agent powered by the Muse Spark model family. Designed for multi-agentic software engineering, it inspects repositories, plans architectural changes with approval gating, executes edits across multiple files, and maintains crash-resilient session event logs.

Is Muse Code free?

No — Muse Code is a paid tool. Muse Code is billed either pay-as-you-go through a Meta Model API key or with a flat-rate Muse Code subscription. Pay-as-you-go Muse Spark rates per 1M tokens: Standard mode $1.25 input, $0.15 cached input, $4.25 output (content not used for training); Contributor mode $0.10 input, $0.002 cached input, $0.20 output in exchange for letting Meta train on your content. Muse Wonder (Standard only) costs $4.00 input, $0.20 cached input and $20.00 output per 1M tokens. Web search grounding is $2.50 per 1,000 queries. Subscriptions come in four tiers (Everyday, High, Power and Maximum Usage); prices are shown during onboarding and are not published in the docs, and subscriptions default to Contributor mode, which can be turned off.

Is Muse Code still maintained?

Yes — Muse Code is active. Its listing was last verified on October 16, 2026.

What are the best Muse Code alternatives?

The first editor-selected Muse Code alternatives are Claude Code, Codex, Factory Droid, and more.