The Cloudflare API MCP server now includes the Cloudflare Skills catalog through the Skills‑over‑MCP extension. By exposing skills via the standard MCP endpoint, any client that implements the extension can enumerate and retrieve skill assets without building a separate API call.
How Skills Are Exposed
When the extension is active, the server advertises a skills/list operation. A client issues this request and receives a list of available skill identifiers. Each skill’s files are then addressable with a URI that follows the skill://<name>/<path> scheme, allowing direct read access to the underlying resources.
Client Configuration Steps
To consume the new capability, add the Cloudflare MCP endpoint to the client configuration:
https://mcp.cloudflare.com/mcpThe client must also declare support for the Skills‑over‑MCP extension; otherwise the skills/list operation will be unavailable. After the endpoint is added, the workflow is:
- Call
skills/listto retrieve the catalog. - Select a skill name from the response.
- Read required files using the
skill://<name>/<path>URI.
Operational and Architectural Implications
Integrating the skill catalog into existing MCP‑based pipelines removes the need for a separate skills service. Architects should consider the following:
- Endpoint addition is a single configuration change; no additional network paths are introduced.
- Clients must be updated to support the extension, which may involve library version bumps.
- Skill files are fetched on demand, so latency and bandwidth usage will reflect the size and frequency of skill access.
- Because the server now serves both API calls and skill assets, monitoring should include both request types to detect abnormal traffic patterns.
Related CloudNinjas coverage: AI engineering.
What This Means For Practitioners
AI engineers can pull skill definitions directly into model‑context pipelines, cloud and platform teams can embed skills in automation without extra services, and SREs gain a single point of observability for skill‑related traffic. The immediate action is to verify that your MCP client supports the Skills‑over‑MCP extension and to add the Cloudflare MCP endpoint to your configuration. After deployment, track skills/list and skill:// request metrics to ensure expected performance and to spot any unexpected usage.
