TeamRetro MCP Server for LLM Retrospectives and Reports

Overview

The TeamRetro MCP Server implements a Model Context Protocol (MCP) adapter that lets LLMs interact with TeamRetro data and workflow objects programmatically. It exposes managed resources such as users, teams, members, retrospectives, health checks, actions, agreements and reports through a set of HTTP endpoints and tool-like interfaces suitable for LLM-driven agents. The result is an LLM-accessible backend that can create, update, query and summarize TeamRetro artifacts as part of automated retrospectives, reporting pipelines and assistant workflows.

This server is useful when you want to build a conversational agent or automation that can operate on TeamRetro on behalf of users—e.g., creating a retrospective from a conversational prompt, generating a summary report, or reconciling action items. By following MCP conventions, the server makes it straightforward to present these capabilities to any LLM that supports tool calling or external function invocation.

Features

• Exposes core TeamRetro resources as MCP-style tools/endpoints: users, teams, members, retrospectives, health checks, actions, agreements, reports
• CRUD operations for retrospectives and associated items
• Endpoints to fetch and export retrospective reports and summaries
• Health check APIs for monitoring and service-level diagnostics
• Designed to be used by LLMs or agent frameworks that support external tool invocation
• Configurable via environment variables and runnable locally or in containers

Installation / Configuration

Clone the repository and install dependencies (example assumes a Node.js based project):

# Clone
git clone https://github.com/adepanges/teamretro-mcp-server.git
cd teamretro-mcp-server

# Install dependencies (Node.js example)
npm install

# Run locally
npm start

Common environment variables you should configure (example names — check the repo for exact names):

PORT=3000
TEAMRETRO_API_URL=https://api.teamretro.io
TEAMRETRO_API_KEY=your_teamretro_api_key
DATABASE_URL=postgres://user:password@localhost:5432/teamretro
LOG_LEVEL=info

Docker usage:

# Build
docker build -t teamretro-mcp-server .

# Run
docker run -e TEAMRETRO_API_KEY=$TEAMRETRO_API_KEY -p 3000:3000 teamretro-mcp-server

Available Resources

The server exposes resource endpoints suitable for LLM tool invocation. Below is a representative mapping (URI and operation summary). Confirm exact routes in the repository documentation.

Example: Create a retrospective (curl):

curl -X POST "http://localhost:3000/mcp/retros" \
  -H "Authorization: Bearer ${TEAMRETRO_API_KEY}" \
  -H "Content-Type: application/json" \
  -d '{
    "teamId": "team_123",
    "title": "Sprint 42 Retro",
    "startDate": "2026-04-01",
    "endDate": "2026-04-07",
    "facilitatorId": "user_456"
  }'

Example: Fetch a generated report:

curl "http://localhost:3000/mcp/retros/retro_789/reports/latest" \
  -H "Authorization: Bearer ${TEAMRETRO_API_KEY}"

Use Cases

• Automating retro creation from chat: An LLM assistant can take a plain-language request (“Create a retro for Sprint 42 with Alice as facilitator”) and call the retros creation tool to provision the meeting and notify members.
• Generating and summarizing reports: After a retrospective, an agent can call the reports endpoint to fetch a summary and then synthesize action items and a one-paragraph executive summary for stakeholders.
• Action item triage and follow-up: A scheduled agent can list open actions across retros, update statuses, and post reminders to team members or create follow-up retrospectives automatically.
• Health and monitoring: Integrate MCP health endpoints with observability tooling to surface service-level issues to ops or prompt an LLM assistant to diagnose misconfigurations.
• Member synchronization: Keep user directories synchronized by having an LLM-driven process call members and teams endpoints to reconcile external directory changes with TeamRetro.

Getting Started Tips

• Read the repository README and inspect the route definitions to confirm exact endpoint names and required payloads.
• Secure the server and API keys: use secrets management for production and limit token scopes where possible.
• When integrating with an LLM, map MCP endpoints to tool names your agent framework understands and provide concise tool descriptions and schemas so the model can reliably call them.

For the canonical source and implementation details, see the project on GitHub: https://github.com/adepanges/teamretro-mcp-server.