Cloudflare is a Cloud Platforms MCP server that lets Claude Code, Cursor, Windsurf and any MCP-compatible AI agent workers, KV, R2, D1, DNS management. Install in 1 minute with mcpizy install cloudflare.
Cloud Platforms
Workers, KV, R2, D1, DNS management
Official homepagemcpizy install cloudflarenpx -y @cloudflare/mcp-server-cloudflarelist_accountsList Cloudflare accounts
list_zonesList zones (domains) on the account
list_workersList deployed Workers
get_worker_logsFetch logs from a Worker
Inputs
worker_namestringrequiredlist_r2_bucketsList R2 storage buckets
list_kv_namespacesList Workers KV namespaces
search_docsSearch Cloudflare documentation
Inputs
querystringrequiredWorks identically across clients. Only the config file path differs.
~/.claude.json{
"mcpServers": {
"cloudflare": {
"command": "npx",
"args": [
"-y",
"mcp-remote",
"https://observability.mcp.cloudflare.com/sse"
]
}
}
}.cursor/mcp.json{
"mcpServers": {
"cloudflare": {
"command": "npx",
"args": [
"-y",
"mcp-remote",
"https://observability.mcp.cloudflare.com/sse"
]
}
}
}~/.codeium/windsurf/mcp_config.json{
"mcpServers": {
"cloudflare": {
"command": "npx",
"args": [
"-y",
"mcp-remote",
"https://observability.mcp.cloudflare.com/sse"
]
}
}
}Authenticate via OAuth on first connect (browser-based)
Paste any of these prompts into Claude Code, Cursor or another MCP-compatible client.
“List my Workers and tell me which ones errored in the last hour”
Uses: list_workers, get_worker_logs
“Show me every R2 bucket in my account”
Uses: list_r2_buckets
“Find Cloudflare docs about configuring D1 databases”
Uses: search_docs
Cloudflare MCP covers the broader Cloudflare platform — Workers, R2, KV, D1, Pages, DNS, Zero Trust. The official MCP from Cloudflare ships several focused servers (one per product) rather than a single monolith, which keeps the tool count per agent manageable. Authentication is a Cloudflare API token scoped to specific resources; the platform's permission model is granular enough that you can give the agent read-only DNS and write Workers without leaking anything else.
We use Cloudflare MCP for three workflows. First, DNS triage: a domain stops resolving, the agent walks records via `list_dns_records`, checks proxied vs DNS-only, and fixes the misconfigured row. Second, Workers debugging: `tail_worker` streams live logs from a deployed Worker — invaluable for debugging an edge function in production. Third, R2 cost analysis: the agent lists buckets, sums sizes, and surfaces buckets eating quota. Token cost varies widely by product; DNS record lists are tiny (~500 tokens), Worker tail can run thousands of tokens per minute if the Worker is chatty.
Compared to AWS MCP (when it ships) or GCP MCP, Cloudflare wins on edge-first primitives and on simplicity — fewer concepts, faster API. The honest trade-off: the per-product MCP split means the agent has to know which server to use for a given task, and the line between "Workers" and "Pages Functions" is blurry. For comprehensive multi-product workflows, you'll often need 3-4 Cloudflare MCP servers loaded simultaneously.
API tokens have a granular scope model — "Account:Cloudflare Workers:Edit" is a different permission from "Zone:Workers Routes:Edit." An agent failing with a 403 usually needs a more specific scope, not a broader one.
Worker logs via `tail_worker` are streaming-only — there's no historical query. If you need to debug an error that happened an hour ago, you need to have had `tail` running at the time, or pre-configure Logpush to a log sink.
D1 databases via MCP support raw SQL but the connection model is per-database. The agent has to know the D1 database ID upfront; `list_databases` then pin the ID before any `execute_sql` call.
DNS changes propagate fast on Cloudflare (seconds to minutes) but the MCP doesn't wait for propagation. If the agent updates a record and immediately tests resolution, the local resolver may return cached. Have the agent advise the user about DNS TTLs.
Honest pros/cons against the closest cloud platforms MCP servers.
| Server | Strengths | Trade-offs |
|---|---|---|
| AWS Route 53 MCP (community) | Right pick if your entire stack is AWS | No edge network, no Workers-equivalent runtime |
| Vercel MCP | Simpler for app-platform use cases | No DNS / Zero Trust / R2 — narrow app-platform scope |
| DigitalOcean MCP (community) | Simpler pricing, predictable for indie devs | No edge network, smaller global footprint |
If Cloudflare doesn't fit your stack, these Cloud Platforms MCP servers solve similar problems.
The Cloudflare MCP server is an Cloud Platforms Model Context Protocol server that lets Claude Code, Cursor, Windsurf, VS Code with Copilot, and other MCP-compatible AI agents workers, KV, R2, D1, DNS management. It exposes Cloudflare's capabilities as tools the AI can call directly from your editor or CLI.
The fastest way is the MCPizy CLI: run `mcpizy install cloudflare` and MCPizy will add the server to your `.claude.json` automatically. You can also install it manually by adding an entry under `mcpServers` in `.claude.json` with the command `npx -y @cloudflare/mcp-server-cloudflare` and restarting Claude Code.
Yes. The Cloudflare MCP server is free and open source (see the GitHub repository linked on this page). You may still need a Cloudflare account or API key to connect the server to the underlying service, but the MCP layer itself has no MCPizy subscription cost.
Yes. Any MCP-compatible client works — including Claude Code, Claude Desktop, Cursor (via `.cursor/mcp.json`), Windsurf, VS Code with Copilot Chat, and custom agents built on the MCP SDK. The same install command targets all of them; only the config file path differs.
Once installed, your AI agent can workers, KV, R2, D1, DNS management directly inside your conversation. Typical use cases include asking Claude Code or Cursor to run Cloudflare operations, inspect results, chain Cloudflare with other MCP servers (see our Workflow Recipes), and automate repetitive cloud platforms tasks without leaving your editor.