mcp

This MCP server brings AI-powered management capabilities for Fastly into your environment. It exposes a Model Context Protocol (MCP) interface that AI assistants can connect to in order to securely list services, view domains, deploy changes, analyze performance, and troubleshoot issues with Fastly configurations. The server supports a set of built-in tools described in the repository documentation, including commands to list available actions, describe resources, execute operations, and query results from cache and performance analyses. It also provides output handling features such as pagination and truncation to keep responses readable during long operations. You can run the server via a container (recommended) or install a native build if you prefer a local binary, then connect your MCP client to the server endpoint to start conversing with your AI agent about your Fastly setup.

Prerequisites:

• Docker installed and running, or a local Go environment if you choose to build from source
• Access to Fastly account and authenticated CLI if you plan to perform actions that require authentication

Option A: Run via Docker (recommended)

1. Install Docker on your host
2. Pull and run the MCP server image: docker pull fastly/mcp:latest docker run -i fastly/mcp:latest
3. The container will start the MCP server and expose its MCP interface to connect your AI assistant.

Option B: Build from Source (Go)

1. Ensure Go 1.20+ is installed
2. Clone the repository: git clone https://github.com/fastly/mcp.git cd mcp
3. Build the binary: go build ./cmd/fastly-mcp
4. Run the binary: ./fastly-mcp

Option C: Pre-built binaries (if available)

1. Download the appropriate binary for your platform from the releases page
2. Make it executable and run: chmod +x fastly-mcp ./fastly-mcp

Environment considerations:

• If using Docker, you may need to pass environment variables for Fastly authentication tokens or API endpoints as described in the project documentation. Prepare a secure method to provide API tokens (e.g., environment variables or mounted secret files).
• The MCP server supports multiple transport modes (StdIO, HTTP, Server-Sent Events). Choose the transport that best fits your orchestration or agent integration workflow.
• For production deployments, consider enabling security features such as command allowlists, input validation, and prompt injection protections as outlined in the documentation.
• If you encounter connectivity issues with the MCP endpoint, verify network policies, container health, and that the agent is pointed at the correct MCP address/port.