MCP Server

The CI/CD Watch MCP server lets AI assistants query your pipeline data directly. Ask questions like “which pipelines are slowest?” or “what are my DORA metrics?” and get answers grounded in your CI/CD data.

What is MCP?

The Model Context Protocol (MCP) is an open standard that lets AI assistants connect to external data sources. When you add the CI/CD Watch MCP server, your assistant gains access to tools for querying pipeline runs, metrics, and costs.

Supported Clients

Any MCP-compatible client works, including:

Claude Desktop and Claude Code
Cursor
Windsurf
GitHub Copilot in VS Code (1.101+), Visual Studio, JetBrains IDEs, and Copilot CLI
Any other client that supports MCP HTTP transport

Setup

1. Create an API key

Go to Settings > API Keys and create a key with readscope. Copy the key, you'll need it in the next step.

2. Add to your MCP client

Add the config block below to your MCP client. The file location and the JSON top-level key depend on the client.

Claude Desktop, Claude Code, Cursor, Windsurf

Config-file paths:

Claude Desktop: ~/Library/Application Support/Claude/claude_desktop_config.json
Claude Code: ~/.claude.json or project .mcp.json
Cursor: .cursor/mcp.json in your project

{
  "mcpServers": {
    "cicd-watch": {
      "type": "http",
      "url": "https://mcp.cicd.watch",
      "headers": {
        "Authorization": "Bearer cw_your_key_here"
      }
    }
  }
}

GitHub Copilot (VS Code, JetBrains, Visual Studio)

VS Code supports remote HTTP MCP servers from version 1.101 onwards. The config-file path is .vscode/mcp.json in your project, or your user settings.json for a global install. The JSON shape uses servers rather than mcpServers:

{
  "servers": {
    "cicd-watch": {
      "type": "http",
      "url": "https://mcp.cicd.watch",
      "headers": {
        "Authorization": "Bearer ${input:cicd_watch_api_key}"
      }
    }
  },
  "inputs": [
    {
      "id": "cicd_watch_api_key",
      "type": "promptString",
      "description": "CI/CD Watch API key",
      "password": true
    }
  ]
}

The inputs array prompts for the API key at runtime instead of storing it in the config file. JetBrains IDEs and Visual Studio follow the same shape under their respective Copilot MCP settings.

Replace cw_your_key_here with your API key. Nothing to install, the MCP server runs on our infrastructure.

Available Tools

Once connected, your AI assistant can use these tools:

Tool	Description
list-runs	List recent pipeline runs, optionally filtered by time period
list-connections	List all CI/CD provider connections
get-dora-metrics	Get the five DORA metrics (Deployment Frequency, Change Lead Time, Change Fail Rate, Failed Deployment Recovery Time, Deployment Rework Rate)
get-costs	Get cost breakdown including compute costs, wait time, and waste analysis
get-performance	Get pipeline performance analysis with duration stats, trends, and suggestions
list-audit-runs	List CI/CD audit runs in reverse chronological order, with status (pending, running, completed, errored), timing, and worker information
get-audit-run	Get a single CI/CD audit run by ID, including lifecycle timestamps and any error message
list-audit-findings	List CI/CD audit findings filtered by state, pillar, rule, organisation, or repository. Free-tier callers receive counts only; paid tiers receive the full finding list with evidence

The four analytical tools (list-runs, get-dora-metrics, get-costs, get-performance) accept an optional periodDaysparameter (1–365). Defaults to 30 days for metrics tools. The audit tools take their own inputs: get-audit-run takes a run ID; list-audit-findings takes optional state / pillar / rule / org / repo filters. See the Audit docs for the underlying concepts.

Example Prompts

Once configured, try asking your AI assistant:

>“What are my DORA metrics for the last 30 days?”
>“Which pipelines are the most expensive?”
>“Show me recent failed builds”
>“How is our deployment frequency trending?”
>“What optimization opportunities do we have?”