ui-automation-control-plane

MCP Quickstart (1 Pager)

Use This Page For

• Fast execution when you do not want to bounce between MCP docs.
• A clear default path plus failure branches.
• Getting to a result first, then drilling into advanced/internal docs.
• Running the same read-mostly loop from Codex, Claude Code, or another MCP-capable client without pretending Prooflane is a generic assistant.

Installation Truth

• Repo-native start today: pnpm mcp:start
• Publish-ready package target: @uiq/mcp-server
• Planned CLI / bin name: prooflane-mcp
• Protocol today: stdio only
• Local stdio startup does not use OAuth
• npx -y @uiq/mcp-server / pnpm dlx @uiq/mcp-server is the truthful package-launch shape, but it is not published yet

Layering Map

• Default: this page + the standard path in docs/how-to/mcp-clients-setup.md.
• Advanced: quality analysis, automation APIs, and proof campaigns.
• Internal: the full tool inventory and runOverrideSchema contract in docs/mcp.md.

Core 12 Tools (Recommended Working Set)

1. uiq_server_selfcheck
2. uiq_catalog
3. uiq_backend_runtime
4. uiq_api_sessions
5. uiq_api_flows
6. uiq_api_templates
7. uiq_api_runs
8. uiq_run_profile
9. uiq_run_stream
10. uiq_run_overview
11. uiq_read_artifact
12. uiq_gate_failures

Run Lane Map

• uiq_api_runs = workflow /api/runs lane
• uiq_api_automation_run = automation command /api/automation/run lane
• uiq_run_profile / uiq_run_stream = governed pnpm uiq run lane

Treat these as separate surfaces. They are connected, but they are not the same run contract.

Typical Call Chains

1) Default Deep-Load (Most Common)

uiq_server_selfcheck -> uiq_run_profile -> uiq_run_stream -> uiq_run_overview -> uiq_read_artifact

2) Manual Breakdown (Need Fine Control)

uiq_backend_runtime -> uiq_server_selfcheck -> uiq_run_stream -> uiq_run_overview -> uiq_read_artifact

3) Quality Loop (Audit + Autofix + Rerun)

uiq_run_stream -> uiq_run_overview -> uiq_gate_failures -> uiq_read_manifest

4) Release Review Loop (Wave 2)

uiq_run_proof_campaign -> uiq_generate_release_brief -> uiq_read_run_ai_review -> uiq_find_similar_failures

5) Manual Gate Triage Loop (Wave 2)

uiq_run_overview -> uiq_list_manual_gates -> uiq://manual-gates/inbox-summary

6) Codex / Claude Code Review Loop

uiq_run_overview -> uiq_generate_release_brief -> uiq_find_similar_failures -> uiq_list_manual_gates -> uiq_explain_template_feasibility

Use this when an MCP-capable client needs a fast answer to:

• What happened in the latest run?
• Is there a similar historical failure?
• Is a paused run waiting on a manual gate?
• Will this reusable journey drift when it moves to another target family?

Failure Branches (Fast Decision)

• If uiq_server_selfcheck fails:
  • Stop run pipeline.
  • Fix runtime/services/api/token wiring.
  • Re-run uiq_server_selfcheck.
• If uiq_run_overview.gateStatus != "passed":
  • Call uiq_gate_failures.
  • Read evidence via uiq_read_artifact (from failed check paths).
  • Optional deep drill with uiq_read_manifest.
• If artifact path is unclear:
  • Use uiq_run_overview first to confirm current/last run context.
  • If you enabled advanced group, use uiq_list_runs + uiq_read_manifest for deep trace.
• If run behavior needs stronger control:
  • Switch to manual chain (uiq_run_profile -> uiq_run_stream -> uiq_run_overview).

Minimal Inputs To Remember

• Default env (client side):
  • UIQ_MCP_PERFECT_MODE=true
• Backend lane:
  • UIQ_MCP_API_BASE_URL=http://127.0.0.1:18080
• Optional groups (only when needed):
  • UIQ_MCP_TOOL_GROUPS=advanced,register,proof,analysis
• Provider policy: Gemini-only.

Minimal Config Seed

Repo-native today

{
  "mcpServers": {
    "uiq": {
      "command": "pnpm",
      "args": ["mcp:start"],
      "cwd": "/ABSOLUTE/PATH/TO/REPO",
      "env": {
        "UIQ_MCP_API_BASE_URL": "http://127.0.0.1:18080",
        "UIQ_MCP_TOOL_GROUPS": "advanced,analysis,proof"
      }
    }
  }
}

Publish-ready package shape (not live yet)

{
  "mcpServers": {
    "uiq": {
      "command": "npx",
      "args": ["-y", "@uiq/mcp-server"],
      "env": {
        "UIQ_MCP_API_BASE_URL": "http://127.0.0.1:18080",
        "UIQ_MCP_TOOL_GROUPS": "advanced,analysis,proof"
      }
    }
  }
}

Next Docs

• Agent loop walkthrough: docs/how-to/mcp-agent-review-loop.md
• Setup details: docs/how-to/mcp-clients-setup.md
• Contract details: docs/mcp.md

This site is open source. Improve this page.