CI/CD MCP Server

Give your AI agents visibility into CI/CD pipelines. This MCP server connects to GitHub Actions so agents can check build status, read logs, download artifacts, trigger deployments, and re-run failed jobs — all through natural language.

What It Does

When your agent needs to know "why did the build fail?" or "is it safe to deploy?", it calls these tools to get real-time CI/CD data instead of guessing.

Architecture

Tools (8)

Verified Output

Tested against a real repository with GitHub Actions:

> list_pipelines(owner: "zavora-ai", repo: "computer-use-mcp")

{
  "count": 15,
  "runs": [
    { "id": 25970003272, "name": "CI — Build & Test", "conclusion": "success", "branch": "main" },
    { "id": 25969979497, "name": "CI — Build & Test", "conclusion": "success", "branch": "chore/bump-6.2.0" },
    { "id": 25969865660, "name": "CI — Build & Test", "conclusion": "failure", "branch": "main" }
  ]
}

> get_pipeline_run(owner: "zavora-ai", repo: "computer-use-mcp", run_id: 25970003272)

{
  "name": "CI — Build & Test",
  "status": "completed",
  "conclusion": "success",
  "jobs": [
    { "name": "Build native — macOS x64", "conclusion": "success" },
    { "name": "Build native — macOS arm64", "conclusion": "success" },
    { "name": "Build native — Windows x64", "conclusion": "success" },
    { "name": "Build native — Linux x64", "conclusion": "success" },
    { "name": "Test — Linux x64", "conclusion": "success" }
  ]
}

> get_job_logs(owner: "zavora-ai", repo: "computer-use-mcp", run_id: 25970003272)

{
  "jobs": [{
    "name": "Build native — macOS x64",
    "conclusion": "success",
    "steps": [
      { "name": "Set up job", "conclusion": "success" },
      { "name": "Run actions/checkout@v4", "conclusion": "success" },
      { "name": "Install Rust toolchain", "conclusion": "success" },
      { "name": "Install Linux dependencies", "conclusion": "skipped" },
      { "name": "Rust cache", "conclusion": "success" },
      { "name": "Build native module", "conclusion": "success" }
    ]
  }]
}

> download_ci_artifact(owner: "zavora-ai", repo: "computer-use-mcp", run_id: 25970003272)

{
  "count": 4,
  "artifacts": [
    { "name": "computer-use-napi.win32-x64.node", "size_in_bytes": 550209 },
    { "name": "computer-use-napi.darwin-arm64.node", "size_in_bytes": 602482 },
    { "name": "computer-use-napi.linux-x64.node", "size_in_bytes": 667435 }
  ]
}

Installation

1. Build

git clone https://github.com/zavora-ai/mcp-cicd
cd mcp-cicd
cargo build --release

2. Set your GitHub token

export GITHUB_TOKEN=ghp_xxxxxxxxxxxx

You need a token with repo and workflow scopes. Create one at https://github.com/settings/tokens.

3. Add to your MCP client

Claude Desktop — add to ~/Library/Application Support/Claude/claude_desktop_config.json:

{
  "mcpServers": {
    "cicd": {
      "command": "/path/to/mcp-cicd",
      "env": { "GITHUB_TOKEN": "ghp_xxxx" }
    }
  }
}

Kiro — add to .kiro/settings/mcp.json:

{
  "mcpServers": {
    "cicd": {
      "command": "/path/to/mcp-cicd",
      "env": { "GITHUB_TOKEN": "ghp_xxxx" }
    }
  }
}

Cursor — add to .cursor/mcp.json:

{
  "mcpServers": {
    "cicd": {
      "command": "/path/to/mcp-cicd",
      "env": { "GITHUB_TOKEN": "ghp_xxxx" }
    }
  }
}

4. Use it

Ask your agent things like:

• "What's the CI status for main?"
• "Why did the last build fail?"
• "Re-run the failed build"
• "What artifacts were produced by the latest successful build?"
• "Deploy main to staging"

How It Works

The server connects to the GitHub Actions API using your personal access token. When your agent calls a tool, the server makes the appropriate API call and returns structured data the agent can reason about.

Your Agent → MCP Tool Call → mcp-cicd → GitHub Actions API → Response

No data is stored locally. Every call is a live query against GitHub.

Governance

MCP Server Manifest

server_id = "mcp_cicd"
display_name = "CI/CD MCP"
version = "1.0.0"
domain = "developer"
risk_level = "medium"
writes_allowed = "gated"
transports = ["stdio"]
governance_gates = ["deployment_requires_promotion_policy"]

Contributors

James Karanja Maina

License

Apache-2.0 — see LICENSE for details.

Part of the ADK-Rust Enterprise MCP server ecosystem.

Registry Compliance

This server implements the ADK MCP SDK contract:

• HealthCheck — async health probe for registry monitoring
• mcp-server.toml — manifest declaring tools, risk classes, and credentials
• Structured tracing — RUST_LOG env-filter for observability