Cloudflare MCP Server

Exposes Cloudflare platform APIs as discoverable MCP tools through a primary HTTP endpoint with streamble-http streaming transport, enabling LLM clients to invoke functions with structured schemas. The architecture uses a standardized tool registry pattern where each server declares available tools with JSON schemas, parameter definitions, and execution handlers that the MCP protocol can introspect and invoke. This differs from direct API consumption by providing a protocol-agnostic abstraction layer that normalizes authentication, error handling, and response formatting across 15+ specialized servers.

Implements both HTTP streaming (/mcp) and legacy Server-Sent Events (/sse) transport mechanisms with pluggable authentication supporting OAuth 2.0 flows for user-based access and API token mode for programmatic access. The authentication layer uses Cloudflare's identity infrastructure to validate credentials, establish user context, and manage session state across stateless Workers deployments. Each server instance validates incoming requests against the authentication provider before exposing tools, ensuring that only authorized users can invoke Cloudflare operations.

Exposes Cloudflare Audit Logs operations through MCP tools for querying account activity, generating compliance reports, and monitoring security events. The Audit Logs Server implements tools for filtering logs by action type, actor, timestamp, and resource, enabling LLM agents to investigate security incidents and generate audit trails without direct access to log systems. This capability integrates with Cloudflare's audit infrastructure to provide searchable, structured logs of all account operations.

Exposes Cloudflare DNS Analytics operations through MCP tools for querying DNS query patterns, analyzing traffic by geography and query type, and identifying DNS-based threats. The DNS Analytics Server implements tools for retrieving aggregated DNS metrics, understanding query patterns, and detecting anomalies. This capability enables LLM agents to analyze DNS traffic and understand domain usage patterns without direct access to analytics infrastructure.

Exposes Cloudflare Logpush operations through MCP tools for configuring log datasets, managing log destinations, and retrieving streaming logs. The Logpush Server implements tools for setting up log delivery to external systems, querying available log datasets, and retrieving structured logs for analysis. This capability enables LLM agents to configure logging infrastructure and access logs without direct access to Logpush configuration systems.

Provides reusable infrastructure packages (@repo/mcp-common, @repo/mcp-observability, @repo/eval-tools) that all 15+ MCP servers depend on for authentication, metrics collection, and testing. The monorepo uses pnpm workspaces and Turbo for dependency management and build orchestration, enabling consistent tool schemas, error handling, and observability across all servers. This architecture allows new MCP servers to be added without duplicating authentication or metrics logic.

Deploys 15+ MCP servers as Cloudflare Workers at dedicated subdomains (*.mcp.cloudflare.com) with automatic scaling, failover, and edge-based request routing. The deployment architecture uses Wrangler for Worker configuration and deployment, with environment-specific settings for development, staging, and production. Each server instance is stateless and horizontally scalable, with shared state managed through Durable Objects and KV storage.

Exposes Cloudflare Workers runtime metrics, logs, and execution traces through MCP tools that query the Workers Analytics Engine and Logpush APIs. The Workers Observability Server implements tools for retrieving request metrics, error rates, CPU time, and structured logs from deployed Workers, enabling LLM agents to diagnose performance issues and understand runtime behavior without direct API calls. This capability integrates with Cloudflare's native observability stack (Analytics Engine, Logpush, tail logs) to provide real-time and historical insights into Worker execution.

Exposes Cloudflare Workers KV storage operations through MCP tools for reading, writing, listing, and deleting key-value pairs across namespaces. The implementation uses the Cloudflare API to perform KV operations on behalf of the authenticated user, enabling LLM agents to manage application state, cache data, and persist configuration without direct KV API access. Tools support batch operations, metadata queries, and namespace enumeration, allowing agents to understand and manipulate the full KV storage structure.

Exposes Cloudflare R2 (S3-compatible object storage) operations through MCP tools for uploading, downloading, listing, and deleting objects across buckets. The implementation uses the Cloudflare API to perform R2 operations, enabling LLM agents to manage files, artifacts, and large data objects without direct S3 API access. Tools support multipart uploads for large files, bucket enumeration, object metadata queries, and signed URL generation for sharing access.

Exposes Cloudflare D1 (SQLite-based distributed database) operations through MCP tools for executing SQL queries, retrieving schema information, and managing database connections. The implementation uses the Cloudflare API to execute queries against D1 databases, enabling LLM agents to perform data analysis, schema exploration, and CRUD operations without direct database access. Tools support parameterized queries to prevent SQL injection, transaction management, and schema introspection for understanding database structure.

Exposes Cloudflare DNS management operations through MCP tools for creating, updating, deleting, and querying DNS records across zones. The implementation uses the Cloudflare API to perform DNS operations, enabling LLM agents to manage domain configurations, set up routing rules, and configure DNS-based security policies without direct API access. Tools support all DNS record types (A, AAAA, CNAME, MX, TXT, etc.), zone enumeration, and record filtering.

Exposes Cloudflare Workers Builds operations through MCP tools for triggering deployments, monitoring build status, and retrieving deployment history. The Workers Builds Server implements tools for querying build logs, understanding deployment status, and triggering new builds through the MCP interface. This capability integrates with Cloudflare's CI/CD infrastructure to enable LLM agents to manage the full deployment lifecycle without direct access to build systems.

Exposes Cloudflare AI Gateway operations through MCP tools for configuring request routing, managing model endpoints, and monitoring API usage across multiple LLM providers. The AI Gateway Server implements tools for setting up rate limiting, failover policies, and request transformation rules that control how LLM requests are routed through Cloudflare's infrastructure. This capability enables LLM agents to manage their own routing configuration and understand request patterns without direct access to gateway configuration systems.

Exposes Cloudflare AutoRAG operations through MCP tools for indexing documents, managing vector embeddings, and performing semantic search across knowledge bases. The AutoRAG Server implements tools for uploading documents, querying indexed content, and retrieving relevant context for LLM prompts. This capability enables LLM agents to build and query knowledge bases without direct access to vector storage or embedding infrastructure.