TechCrunch · AIAug 6, 2026, 05:21 UTC+8

Meta launches Muse Code, a terminal coding agent for large code bases

Meta launches Muse Code, an AI agent for large code bases

Meta released Muse Code in beta, a terminal coding agent powered by its Muse Spark model. It handles planning, coding, and validation across large repos, spawning parallel sub-agents for big jobs without touching your working copy. Meta's AI chief Alexandr Wang told WSJ it could be a strong cost option versus OpenAI Codex and Anthropic Claude Code. The post doesn't disclose pricing or a GA date.

Why it matters: Meta launches a terminal coding agent with concrete mechanisms and direct competitor positioning. Score stays below 80 because it's a beta release with no benchmarks or head-to-head comparisons disclosed — real-world performance remains unverified.

Read the original ↗Export Markdown

Take

This one's worth a click because Meta finally pushed its AI coding tools into terminal agent territory. Muse Code runs on Meta's own Muse Spark model and handles the full loop — planning, coding, validation — across large repos. For big jobs, it spawns parallel sub-agents in isolated workspaces without touching your local code. Meta's AI chief Alexandr Wang told WSJ it could be a strong cost play versus OpenAI Codex and Anthropic Claude Code.

The catch: no pricing, no GA date. The cost advantage is just a quote right now, no numbers to back it up. And the parallel sub-agent approach isn't new — Devin and Claude Code both do something similar. The post doesn't spell out what makes Meta's version different.

If the pricing does come in noticeably lower than Codex and Claude Code, it'd be a natural pick for teams already in Meta's ecosystem. For now, it's a beta — I wouldn't rip out any workflows just yet.

Source: TechCrunch · AI · techcrunch.com
Published: Aug 6, 2026, 05:21 UTC+8 · Aug 6, 2026