
Muse Code Guide
Independent, source-traceable reference for Meta's terminal coding agent
About Muse Code Guide
Muse Code Guide is a documentation site rather than a tool, and it exists because of a genuine gap. When a major vendor ships a coding agent, the available information immediately splits into three unlabelled categories: things that are verifiably true, things the vendor has claimed, and things the community has inferred or assumed. Sorting those apart is real work and almost nobody does it. This site does exactly that for Meta's Muse Code, separating confirmed release facts, Meta's own claims, evaluation limits and open community questions, with sources attached.
Start with the overview for what Muse Code actually is, then the release record for the source-traceable launch facts with their verification dates. The Spark 1.2 section covers the underlying model, and the how-it-works section explains the runtime. Throughout, claims are categorised rather than presented flat, so you can see at a glance whether something is confirmed, asserted by Meta, or an open question. The sources section provides the underlying references, and the method section documents how the distinction is drawn. Because verification dates are shown, you can judge how current any given claim is rather than assuming.
- •Claims Separated by Status - Confirmed facts, vendor claims, evaluation limits and open questions kept distinct rather than blended
- •Source-Traceable Facts - Every claim tied to a source with a stated verification date
- •Stated Evaluation Limits - Explicit about what benchmarks do and do not show, which vendors never are
- •Published Method - Documents how it decides what counts as confirmed, so the editorial standard is inspectable
- •Runtime Documentation - Covers how the agent actually works, not only what it claims to do
For developers evaluating whether to adopt Meta's Muse Code, who want the verified facts separated from the launch marketing before committing time. It suits engineering leads assessing a new agent for a team, where the difference between a demonstrated capability and a claimed one determines the decision. Anyone comparing terminal coding agents gets a source-traceable baseline for one of them.
Pricing
Not published by the vendor
Free to read - an independent reference site with no account, subscription or paywall























































