LabscoMCP early access ↗
pinkroosteraiMCP SERVER

PinkRoosterMcp

Self-hosted project tracking built for coding agents: scaffold a work package in one call, and completing a task cascades state up through phases, work packages and linked issues.

by pinkroosterai ↗Project & Task ManagementCommunity

LABSCO SUMMARY

State cascades are what make it worth wiring into an agent loop.

Completing a task auto-completes its phase when the phase is done, which auto-completes the work package, which resolves the linked issue — and each write returns the downstream changes it caused, so the agent knows the new state without re-reading the project. Access has two lanes: per-project role-based control for humans, and API-key authentication for agents that skips it.

What it is

A project management system designed around an AI agent as the primary user rather than a dashboard user. Work items carry human-readable IDs, responses are compact and carry the state changes a write triggered, and a React dashboard sits alongside for human visibility.

What you get

  • A compact status view and a priority-ordered list of what is actionable, enriched with linked entity names and complexity — get_project_status, get_next_actions
  • Issues and feature requests created, updated and read, with structured user stories managed on feature requests — create_or_update_issue, get_issue_details, get_issue_overview, create_or_update_feature_request, get_feature_request_details, get_feature_requests, manage_user_stories
  • A whole work package created in one call — work package plus phases plus tasks plus dependencies — or built up piece by piece — scaffold_work_package, create_or_update_work_package, get_work_package_details, get_work_packages, create_or_update_phase, create_or_update_task
  • Several task states updated at once with the cascades consolidated into one response, acceptance-criteria verification recorded, and dependencies added or removed with automatic blocking — batch_update_task_states, verify_acceptance_criteria, manage_dependency
  • A persistent knowledge store that survives agent sessions, merged by name so writing to an existing memory appends content and unions tags — list_memories, get_memory_details, create_or_update_memory, delete_memory
  • Sixteen slash-command workflows on top of the tools, including scaffolding from a description, a full implementation loop, and a parallel codebase audit that files the confirmed findings as tracked issues

Requirements

Docker and Docker Compose, and Claude Code if you want the install to register the server for you. No account and no external service — it is self-hosted. Clone the repository and run make install, which writes a working .env, registers the MCP server, installs the workflow skills and pulls the prebuilt image. The dashboard comes up on port 3000, the API on 5100 and MCP on 5200. .NET 9 and Node.js 20+ are needed only for local development.

Setup effort

Build from source — git clone https://github.com/pinkroosterai/PinkRoosterMcp.git

See installation options