Reference — MCP tools¶

Generated from the Knovaryn MCP server's real tool registration (build_server().list_tools()), so names and counts cannot drift from what an MCP client actually sees over stdio or streamable HTTP.

23 registered tools:

Tool	Description
`health`	Health / identity check for the Knovaryn MCP server.
`knovaryn_add_source`	Register an uploaded artifact. Returns source_id + next job handle.
`knovaryn_cancel_job`	Request cooperative cancellation (idempotent).
`knovaryn_compare_runs`	Compare quality/cost across two compatible runs.
`knovaryn_create_dataset_version`	Assemble an immutable draft version from accepted examples.
`knovaryn_create_project`	Create a project. Returns project_id, default profile, and actions.
`knovaryn_doctor`	Environment and dependency capability diagnostics (no secrets).
`knovaryn_estimate_run`	Dry-run estimate for selected sources/topology/profile/targets.
`knovaryn_export_dataset`	Export accepted examples as JSONL; returns artifact handles/summary.
`knovaryn_get_job`	State, stage, progress, counts, costs, warnings, errors, events.
`knovaryn_inspect_source`	Metadata, preflight, extraction summary, license/privacy state.
`knovaryn_license_report`	Source license + privacy report and §15.6 publication-gate decision.
`knovaryn_lineage`	Provenance + lineage for an example (source docs, spans, chunks).
`knovaryn_list_jobs`	Paginated job list (optionally filtered by project).
`knovaryn_list_projects`	Paginated, filtered project summary.
`knovaryn_preview_examples`	Bounded, redacted page of examples + quality dimensions + evidence.
`knovaryn_publish_dataset`	Publish an approved version (external side effect). Dry-run by default.
`knovaryn_resume_job`	Resume a retryable failed/paused job (re-lease and continue).
`knovaryn_review_example`	Accept/reject/relabel an example (records an audit-trail decision).
`knovaryn_run_job`	Execute a queued pipeline job offline in-process (returns final state).
`knovaryn_start_pipeline`	Start a durable pipeline job (offline). Returns a job handle.
`knovaryn_validate_dataset`	Run validation for a project's examples; returns a bounded report.
`run_pipeline`	Run the offline end-to-end pipeline on bundled sample documents.

Run with knovaryn mcp (stdio by default; --transport streamable-http for remote use behind the bearer-token guard). See MCP clients.