Cloudflare

Customization

Before executing, check for user customizations at: ~/.claude/PAI/USER/SKILLCUSTOMIZATIONS/Cloudflare/

If this directory exists, load and apply any PREFERENCES.md, configurations, or resources found there. These override default behavior. If the directory does not exist, proceed with skill defaults.

MANDATORY: Voice Notification (REQUIRED BEFORE ANY ACTION)

You MUST send this notification BEFORE doing anything else when this skill is invoked.

1. Send voice notification:

   curl -s -X POST http://localhost:8888/notify \
     -H "Content-Type: application/json" \
     -d '{"message": "Running the WORKFLOWNAME workflow in the Cloudflare skill to ACTION"}' \
     > /dev/null 2>&1 &
   

2. Output text notification:

   Running the **WorkflowName** workflow in the **Cloudflare** skill to ACTION...
   

This is not optional. Execute this curl command immediately upon skill invocation.

Cloudflare Skill

Deploy and manage Cloudflare Workers, Pages, and services. Uses two complementary tools:

• Code Mode MCP (primary for API operations) — search() + execute() for querying workers, managing KV/R2/D1, checking deployments, DNS, analytics. ~1,069 tokens vs 1.17M for traditional MCP.
• Wrangler (deploy/dev only) — wrangler deploy, wrangler dev, wrangler pages deploy for deploying from local files.

Dual-Mode Reference

Operation	Tool	Why
Deploy Worker	wrangler deploy	Needs local files + wrangler.toml
Deploy Pages	wrangler pages deploy	Needs dist/ directory
Local dev	wrangler dev	Local server
List Workers	MCP execute()	API query, no local files needed
Check deployment status	MCP execute()	API query
Read/write KV	MCP execute()	API operation
Manage DNS	MCP execute()	API operation
View logs	MCP execute()	API query
Inspect R2/D1/Vectorize	MCP execute()	API operations
View analytics	MCP execute()	API query

Workflow Routing

When executing a workflow, output this notification directly:

Running the **WorkflowName** workflow in the **Cloudflare** skill to ACTION...

• Create Worker or MCP server → Workflows/Create.md
• Troubleshoot deployment issues → Workflows/Troubleshoot.md
• Query Cloudflare state (list workers, check KV, DNS, analytics) → Workflows/Query.md

Code Mode MCP Reference

The Cloudflare Code Mode MCP server (https://mcp.cloudflare.com/mcp) exposes the entire Cloudflare API through 2 tools:

search() — Discover endpoints

Find API endpoints by keyword. Use this when you don't know the exact path.

// Example: Find all Workers-related endpoints
search("workers scripts")

// Example: Find KV namespace operations
search("KV namespace")

// Example: Find DNS record endpoints
search("DNS records")

execute() — Call the API

Execute any Cloudflare API endpoint directly. Handles auth automatically via OAuth.

// Example: List all Workers scripts
execute("GET /accounts/{account_id}/workers/scripts")

// Example: Read a KV value
execute("GET /accounts/{account_id}/storage/kv/namespaces/{namespace_id}/values/{key_name}")

// Example: List DNS records
execute("GET /zones/{zone_id}/dns_records")

MCP Auth

Code Mode MCP handles authentication via OAuth — no tokens needed. On first use, it will prompt for Cloudflare login. After that, auth is cached.

Quick Reference

• Account ID: Set via CF_ACCOUNT_ID environment variable
• Worker URL format: https://[worker-name].[your-subdomain].workers.dev

Deployment Commands

Workers Deployment

# Unset tokens that interfere with wrangler login-based auth
(unset CF_API_TOKEN && unset CLOUDFLARE_API_TOKEN && wrangler deploy)

Pages Deployment

CRITICAL: ALL env tokens lack Pages permissions. MUST unset them to use OAuth:

# ALWAYS unset tokens for Pages - OAuth login works, tokens don't
(unset CF_API_TOKEN && unset CLOUDFLARE_API_TOKEN && bunx wrangler pages deploy dist --project-name=PROJECT_NAME --commit-dirty=true)

Critical Notes

• Workers: Unset CF_API_TOKEN and CLOUDFLARE_API_TOKEN before deploying - they interfere with wrangler login-based auth
• Pages: UNSET ALL TOKENS - None of the API tokens have Pages permissions. OAuth-based wrangler login is the ONLY method that works.
• API queries: Use Code Mode MCP instead of manual curl or fetch() to api.cloudflare.com
• Wrangler stays for: deploy, dev, pages deploy, and local config only

Examples

Example 1: Deploy a Worker

User: "deploy the MCP server to Cloudflare"
-> Invokes CREATE workflow
-> Unsets env tokens, runs wrangler deploy
-> Verifies via MCP execute() that worker is live
-> "Deployed to https://mcp-server.[subdomain].workers.dev"

Example 2: Query Cloudflare state

User: "list all my workers"
-> Invokes QUERY workflow
-> MCP search("workers scripts") to find endpoint
-> MCP execute("GET /accounts/{id}/workers/scripts")
-> Returns list of all deployed workers

Example 3: Fix deployment error

User: "Cloudflare deploy is failing with auth error"
-> Invokes TROUBLESHOOT workflow
-> MCP execute() to check deployment status and logs
-> Identifies token interference
-> "Fixed - tokens were overriding OAuth. Redeployed successfully."

Example 4: Check DNS records

User: "what DNS records does example.com have?"
-> Invokes QUERY workflow
-> MCP search("DNS records") + execute() to list records
-> Returns full DNS record table