Cloudflare MCP Server Covers Full API in Two Tools and 1,000 Tokens

Cloudflare

Feb 20, 2026 · Updated Jun 5, 2026

Cloudflare launched a new MCP server that covers its 2,500-endpoint API in two tools and ~1,000 tokens. Code Mode lets agents write code to discover and call APIs instead of loading every tool description upfront, cutting token overhead by 99.9%.

Cloudflare launched a new MCP server covering its full API - DNS, Zero Trust, Workers, R2, and more - using Code Mode, a technique that replaces thousands of tool definitions with just two: search() and execute(). The agent writes JavaScript to discover endpoints and compose API calls, keeping the token footprint fixed at ~1,000 regardless of API size.

The MCP context window problem is real: every tool definition eats tokens, making large APIs practically inaccessible to agents. Code Mode solves this at the architecture level - the agent interacts through code rather than pre-loaded descriptions. Cloudflare is open-sourcing the Code Mode SDK so developers can apply the pattern to their own servers.

If you're building MCP servers for large APIs, this is the pattern to watch. It sidesteps the tool-per-endpoint scaling problem entirely, making any API agent-accessible without blowing through context windows.

View the full update on blog.cloudflare.com

Cloudflare

@CloudflareFeb 20

The Cloudflare API has over 2,500 endpoints. Exposing each one as an MCP tool would consume over 2 million tokens. With Code Mode, we collapsed all of it into two tools and roughly 1,000 tokens of context. https://t.co/rpWBqGao0a

108

View on X

Every HeadsUpAI update is written based on its original source and reviewed before it's published. Read our editorial standards →

See all AI news & updates from Cloudflare →

Keep reading

Cloudflare Launches Docs for Agents to Standardize Machine Readable Documentation

Cloudflare introduced a dedicated documentation hub that provides technical content in agent-friendly formats like Markdown and through the Model Context Protocol. This shift allows AI agents to consume documentation with lower token costs and higher accuracy while gaining direct access to Cloudflare APIs.

Appwrite Launches MCP Server 2.0 to Reduce AI Agent Context Bloat

AppwriteApr 14

Appwrite Launches MCP Server 2.0 to Reduce AI Agent Context Bloat

Appwrite released version 2.0 of its MCP server, replacing dozens of individual tool definitions with a streamlined two-tool architecture. This shift prevents backend service definitions from consuming the model's context window, leaving more room for complex reasoning and code generation.

WebflowMar 22

Webflow MCP Server Now Available in Postman Agent Mode

Webflow's MCP server is now in Postman's Agent Mode, letting developers query and manage sites, CMS collections, and pages via natural language. Ask the agent to fetch your data — it resolves the right API call and returns live results.

falMar 20

fal Launches MCP Server Connecting AI Agents to 1,000+ Generative Models

fal launched an MCP server giving AI assistants like Claude, Cursor, and Windsurf direct access to 1,000+ generative AI models — image, video, audio, and 3D. Assistants handle model discovery and execution automatically from a single conversation turn.