The ultimate harness for coding agents and beyond. All your agents. All your machines. One command center. Start with code, then follow your curiosity and build across disciplines: CAD, circuits, robots, games and music.
TypeScript 3d-modelingagent-frameworkagent-harnessai-agentscadclaude-codecodexcoding-agentsdeveloper-toolsesp32game-developmentmcp-servermulti-agentopen-hardwareopencodeopenharnesspcbroboticsterminaltmux
/**
* Muse Code runtime profile.
*
* Everything the chip needs is in `~/.config/muse/settings.json`, a flat JSON document — no pane
* scraping, no catalog fetch. Verified content (muse 0.1.0-R708.1):
* {"schema_version":1,"provider":"meta","model":"muse-spark-1.2-contributor"}
* `reasoning_effort` is absent until the user changes it; muse's own `--help` documents the default
* as `high`, so an absent key is reported as that rather than as "unknown".
*/
const EFFORTS = new Set(['none', 'minimal', 'low', 'medium', 'high', 'xhigh', 'ultra'])
export function parseMuseSettings(text: string | null): { model: string | null; effort: string | null } {
if (!text) return { model: null, effort: null }
let parsed: unknown
try { parsed = JSON.parse(text) } catch { return { model: null, effort: null } }
const obj = parsed && typeof parsed === 'object' ? parsed as Record<string, unknown> : {}
const model = typeof obj.model === 'string' && obj.model ? obj.model : null
const raw = typeof obj.reasoning_effort === 'string' ? obj.reasoning_effort.toLowerCase() : ''
return { model, effort: EFFORTS.has(raw) ? raw : (model ? 'high' : null) }
}