Home / MCP / Cloudflare MCP Server
Remote MCP servers enabling natural language access to Cloudflare tools and services across observability, bindings, and more.
Configuration
View docs{
"mcpServers": {
"docs": {
"url": "https://docs.mcp.cloudflare.com/mcp"
}
}
}MCP lets you manage Cloudflare services through natural language by connecting to remote MCP servers. You can read configurations, get suggestions, and even apply changes across Cloudflare products using an MCP client. This enables conversational control over services like observability, bindings, and more, across development, security, and performance tooling.
Connect your MCP client to the Cloudflare MCP servers to access a wide range of services. Your client can read configurations from Cloudflare, receive data-driven suggestions, and perform actions across the platform. Use the remote MCP servers by entering their URLs into your client’s MCP server list, then interact with them using natural language prompts. If your client does not support remote MCP servers yet, you can specify them in a local configuration file using a tool like mcp-remote to reference the remote endpoints.
Prerequisites: you need Node.js installed to run MCP client tooling or to use npm/npx-based commands for remote servers.
1) Ensure Node.js is installed. If you don’t have it, install from https://nodejs.org and verify with node -v and npm -v.
2) If you plan to use the remote MCP server approach with a client that doesn’t support remote servers yet, install the mcp-remote helper globally.
3) Add Cloudflare MCP servers to your client. Use the provided URLs for remote access (see the HTTP server entries below) or use the mcp-remote tool to configure them locally.
Configuration and access notes: You can access a broad set of Cloudflare MCP servers via HTTP remote endpoints. The list includes observability, bindings, builds, and many other services, each exposed at a dedicated MCP URL.
Using with OpenAI Responses API: If you want to use a Cloudflare MCP server with the OpenAI responses API, you will need to provide an API token with the required scopes for that server. Create the token in the Cloudflare dashboard with the permissions needed for the specific MCP server you plan to access.
Troubleshooting: If a response is interrupted due to context length, try making queries more specific or splitting a multi-tool chain into smaller requests to avoid hitting limits.
Paid features: Some capabilities may require a paid Cloudflare Workers plan. Ensure your Cloudflare account has the necessary subscription level for the features you intend to use.
Contributing: If you want to contribute or run this setup locally, follow the contribution guidelines and start from a clean environment for reproducible results.
{
"mcpServers": {
"cloudflare-observability": {
"command": "npx",
"args": ["mcp-remote", "https://observability.mcp.cloudflare.com/mcp"]
},
"cloudflare-bindings": {
"command": "npx",
"args": ["mcp-remote", "https://bindings.mcp.cloudflare.com/mcp"]
}
}
}Only access the MCP servers you explicitly configure in your client. Manage API keys and tokens with the minimum scopes required for the tasks you perform. Guard access to remote servers with appropriate credentials and monitor usage to prevent unintended changes.
To start using a downstream MCP server, configure your client to connect to the server’s URL, then begin issuing natural language requests that align with the server’s capabilities, such as querying observability data or deploying bindings-enabled workflows.