@marchese-md/cli

marchese — the command-line tool for MARCHESE.md, the efficiency and orchestration contract layer for AI agent engineering. Validate specs, audit context budgets, route tasks to the cheapest capable model, run orchestration loops, and drive the self-improving knowledge base, all from one binary any script or coding agent can shell out to.

Install

npm install -g @marchese-md/cli
# or run without installing
npx @marchese-md/cli validate MARCHESE.md

Commands

| Command | Description | |---|---| | marchese validate <file> | Validate a MARCHESE.md spec file against the JSON Schema | | marchese audit | Audit workspace context budget, CLAUDE.md lines, skills, and MCP preload | | marchese init | Initialize a new MARCHESE.md specification file | | marchese optimize | Compress prompt payloads (RTK-style) and optionally write output | | marchese gotcha | Manage the gotchas ledger in MARCHESE.md | | marchese loop | Execute a 4-block orchestration loop from MARCHESE.md | | marchese status | Display status line: model tier, effort, context usage, RTK, loops | | marchese route | Decide Codex offload vs local/Claude for a token-heavy task (nuclear route) | | marchese rtk | RTK binary status, bash rewrite, or PreToolUse payload processing | | marchese knowledge | Init/ingest/status for the self-seeded knowledge base (no external dump required) | | marchese improve | Run the self-improving improve-system loop (3 buckets + change-log) |

Every command supports --json for machine-readable output and returns non-zero exit codes on failure, so it's safe to script or wire into a CI gate or an agent's tool-call loop.

Example

marchese validate MARCHESE.md
marchese audit --json
marchese loop run dev-loop --gate

Docs

Full spec: docs/SPEC.md in the main repo.

License

MIT