Skip to content

mcp-tools.md

Latest commit

 

History

History
104 lines (85 loc) · 3.79 KB

mcp-tools.md

File metadata and controls

104 lines (85 loc) · 3.79 KB
Raw

MCP Tool Reference

The @martinloop/mcp package exposes one primary coding execution entrypoint plus planning, inspection, run-control, and review helpers.

Planning and Execution

Tool Purpose
martin_doctor Check local MartinLoop and agent readiness.
martin_plan Outline scope, verification, and budget posture before spending a run.
martin_preflight Validate a proposed run contract before execution.
martin_run Execute a governed coding run. MartinLoop auto-runs the matching readiness checks and only stops when preflight finds a real blocker.

Status and Run Control

Tool Purpose
martin_status Summarize live or recent run status.
martin_logs Read recent run-control receipts and workflow events.
martin_pause Request that a running loop pause at a safe boundary.
martin_cancel Cancel a running loop and capture the control receipt.
martin_continue Continue a paused loop with the recorded run context.

Inspection and Evidence

Tool Purpose
martin_inspect Inspect a persisted loop record.
martin_list_runs List saved run records.
martin_triage_runs Rank saved runs by urgency and missing evidence.
martin_get_run Read one run by ID.
martin_get_attempt Read one attempt from a run.
martin_get_verification_results Read persisted verifier evidence.
martin_run_dossier Produce a compact run dossier.
martin_dossier Alias-first dossier surface for host-friendly receipts.
martin_eval Grade the run for verifier posture, regression risk, and reviewability.

Review Helpers

Tool Purpose
martin_pr_summary Produce a review-ready summary from the run receipts.
martin_create_pr Create a PR when the host explicitly wants GitHub mutation.
martin_review_pr Review an existing PR with MartinLoop evidence context.

Resources

  • martin://server/health
  • martin://runs/recent
  • martin://runs/triage
  • martin://runs/latest
  • martin://runs/latest/summary
  • martin://runs/latest/receipt
  • martin://runs/latest/proof-card
  • martin://runs/latest/budget-status
  • martin://runs/latest/verifier-evidence
  • martin://runs/latest/rollback-evidence
  • martin://policies/current
  • martin://repo/risk-map
  • martin://verifiers/results
  • martin://agent/next-step
  • martin://guides/mcp-usage
  • martin://guides/agent-start
  • martin://guides/command-map
  • martin://guides/ide-onboarding
  • martin://guides/operating-rules
  • martin://guides/publish-readiness

Resource Templates

  • martin://runs/{loopId}
  • martin://runs/{loopId}/dossier
  • martin://runs/{loopId}/attempts/{attemptIndex}
  • martin://runs/{loopId}/verification

Prompts

  • martin_start
  • martin_preflight
  • martin_triage
  • martin_resume
  • martin_prove
  • martin_release_check
  • martin_governed_coding_kickoff
  • martin_debug_failed_run
  • martin_publish_readiness_review
  • martin_triage_run_store
  • safe_bug_fix
  • write_tests_first
  • small_refactor
  • security_review
  • pr_review
  • release_check

Profile Notes

  • minimal is the safest default profile.
  • diagnostic adds deeper run-store and evaluation helpers.
  • full-local adds martin_run plus run-control helpers for local operators.
  • github-review adds PR-oriented helpers when the host explicitly needs them.

The 0.2.7 package line also adds guide resources meant for hosts that want MartinLoop to feel built in rather than bolted on. Start with martin://agent/next-step, martin://guides/command-map, and martin://guides/operating-rules before you open full run JSON.

For installation and host setup, see MCP setup. For platform commitments, see MCP compatibility.