cloudflare-master

name	cloudflare-master
version	1.0.0
description	Cloudflare operations — Workers deploy, DNS, KV, R2, D1, Pages, Tunnels via wrangler CLI and MCP API
author	ninetrix
tags	["cloudflare","serverless","infrastructure"]
requires	{"tools":["wrangler","cloudflare","bash"]}
companion_tool	wrangler

Cloudflare Master

Manage Cloudflare infrastructure using two tools: wrangler CLI for Workers development and the Cloudflare MCP for API operations (DNS, security, analytics).

When this applies

• Deploying or managing Cloudflare Workers
• Managing DNS records, zones, or SSL certificates
• Working with KV, R2, D1, or Queues
• Setting up Cloudflare Tunnels
• Deploying to Cloudflare Pages
• Querying analytics or audit logs

Auth

• wrangler: uses CLOUDFLARE_API_TOKEN env var or wrangler login for OAuth
• MCP: authenticates via OAuth through the MCP server

Workers — wrangler CLI

wrangler init my-worker                       # scaffold new project
wrangler dev                                  # local dev server on :8787
wrangler deploy                               # deploy to production
wrangler deploy --dry-run                     # preview without deploying
wrangler tail                                 # stream live logs
wrangler tail --status error                  # filter to errors only
wrangler secret put SECRET_NAME               # add secret
wrangler secret list                          # list secrets
wrangler delete my-worker                     # remove worker
wrangler rollback                             # revert to previous version
wrangler versions list                        # list deployed versions
wrangler deployments list                     # deployment history

KV (Key-Value Store)

wrangler kv namespace list                    # list namespaces
wrangler kv namespace create MY_KV            # create namespace
wrangler kv key put --namespace-id <id> "key" "value"
wrangler kv key get --namespace-id <id> "key"
wrangler kv key list --namespace-id <id>
wrangler kv key delete --namespace-id <id> "key"
wrangler kv bulk put --namespace-id <id> data.json

R2 (Object Storage)

wrangler r2 bucket list                       # list buckets
wrangler r2 bucket create my-bucket           # create bucket
wrangler r2 object get my-bucket/path/file.txt --file ./local.txt
wrangler r2 object put my-bucket/path/file.txt --file ./upload.txt
wrangler r2 object delete my-bucket/path/file.txt

D1 (SQL Database)

wrangler d1 list                              # list databases
wrangler d1 create my-db                      # create database
wrangler d1 execute my-db --command "SELECT * FROM users LIMIT 10"
wrangler d1 execute my-db --file schema.sql   # run SQL file
wrangler d1 export my-db --output backup.sql  # export

Pages

wrangler pages project list                   # list projects
wrangler pages deploy ./dist                  # deploy directory
wrangler pages deployment list --project my-site

Tunnels

wrangler tunnel list                          # list tunnels
wrangler tunnel create my-tunnel              # create tunnel
wrangler tunnel run my-tunnel                 # run tunnel
wrangler tunnel route dns my-tunnel sub.example.com  # route DNS
wrangler tunnel delete my-tunnel

MCP — Cloudflare API

Use the Cloudflare MCP for operations not covered by wrangler: DNS management, zone settings, firewall rules, analytics, audit logs. The MCP provides search() to find endpoints and execute() to call them.

Decision Rules

• If deploying Workers/Pages → use wrangler CLI
• If managing DNS, zones, firewall, SSL → use Cloudflare MCP
• If need live logs → wrangler tail
• If managing storage (KV/R2/D1) → use wrangler CLI
• If querying analytics or audit logs → use Cloudflare MCP
• If setting up tunnels → use wrangler tunnel

Do

• Use --dry-run before first deploy to preview changes
• Stream logs with wrangler tail after deploying to verify
• Use wrangler secret for sensitive values (never hardcode)
• Check wrangler versions list before rollback

Don't

• Deploy without --dry-run first on unfamiliar projects (may overwrite production)
• Delete workers or tunnels without user confirmation (destructive, affects live traffic)
• Store secrets in wrangler.toml (use wrangler secret put instead — wrangler.toml is committed to git)
• Run wrangler d1 execute with destructive SQL (DROP/DELETE) without confirmation