CLI reference

The maida CLI captures, inspects, compares, and gates agent runs. Commands that accept a trace ID default to the latest run when the ID is omitted; the selected run is announced on stderr so stdout remains machine-readable.

Start and configure

Command	Use it to
maida demo	Run the deterministic first-run or regression story
maida init	Scaffold policy and optional GitHub Actions configuration
maida view	Open the local execution timeline
maida list	List recent local runs

Capture and import

Command	Use it to
maida capture claude-code	Receive local Claude Code OTLP telemetry
maida capture claude-hook	Append one passive Claude Code hook event
maida import claude-code	Normalize a captured Claude Code session
maida import langfuse	Import completed traces through Langfuse’s read-only API
maida validate-trace	Validate an externally emitted native trace
maida export	Write a portable JSON envelope for one run

Baseline and gate

Command	Use it to
maida run	Execute isolated trials and apply a policy-v2 gate
maida baseline	Capture an immutable reviewed baseline
maida accept	Intentionally update a baseline with provenance
maida drift	Evaluate a completed trace window against a baseline
maida scenario run	Gate isolated Claude Code scenarios
maida assert	Evaluate one completed trace through the legacy interface
maida diff	Inspect structural changes between runs or against a baseline
maida extract	Derive review-required policy and baseline drafts