mcp-cloudflare

Slim Cloudflare MCP Server for managing DNS, zones, tunnels, WAF, Zero Trust, and security via Cloudflare API v4.

No SSH. No shell execution. API-only. 3 runtime dependencies.

Table of Contents

• Features
• Quick Start
• HashiCorp Vault Integration (Optional)
• Claude Code Integration
• Configuration
• Multi-Zone Support
• Tools
• Skills
• Development
• License

Features

75 tools across 11 domains:

• DNS — Record management (A, AAAA, CNAME, MX, TXT, SRV, CAA, NS), batch operations
• Zones — Zone listing, settings, SSL/TLS configuration, cache management
• Tunnels — Cloudflare Tunnel creation, configuration, and ingress management
• WAF — Ruleset management, custom firewall rules, rate limiting
• Zero Trust — Access application CRUD (create/delete), policies (create/delete), identity providers (create/delete), Gateway status
• Security — Security event analytics, IP access rules, DDoS configuration, Security Center insights
• Workers KV — Namespace management, key-value read/write/delete, key listing
• Workers — Script deployment, route management
• Worker Secrets — Secret management (names only, values never exposed)
• Worker Analytics — Invocation metrics, CPU time, error rates via GraphQL
• R2 Storage — Bucket management, object listing and metadata, custom domains, location hints

Quick Start

npm install
cp .env.example .env   # Edit with your Cloudflare API token
npm run build
node dist/index.js     # stdio transport for MCP

HashiCorp Vault Integration (Optional)

mcp-cloudflare supports loading Cloudflare credentials from a central HashiCorp Vault instance at startup via AppRole authentication. This is optional — the server works fine with plain environment variables alone.

How It Works

On startup, if NAS_VAULT_ADDR is set the server performs an AppRole login, fetches the KV v2 secret at <mount>/data/cloudflare/api, and injects the values into the process environment before the MCP transport starts. The loader is fully opportunistic:

• If NAS_VAULT_ADDR is unset, the loader is a silent no-op. No Vault calls are made and the server behaves exactly as before.
• On any Vault error (network failure, bad credentials, missing secret path), a single-line warning is written to stderr and the server falls back to whatever environment variables are already set.
• Secret values are never logged. Only the KV path name and a populated-count appear in stderr diagnostics.
• Uses the built-in fetch (Node 20+) — no additional runtime dependencies.

Credential Precedence

Explicit env vars (CLOUDFLARE_API_TOKEN etc.) > Vault > error (missing creds)

If you set CLOUDFLARE_API_TOKEN directly, the Vault loader will not overwrite it. Vault only fills in credentials that are not already present in the environment.

Vault Environment Variables

... View full README on GitHub