HOSTING & DEPLOYMENTS · FOR AI AGENTS
Cloudflare
Operate hosted compute and account resources through the Cloudflare API MCP server.
Use Cloudflare’s remote MCP interface to discover API operations and manage supported resources such as Workers, storage and DNS in an authorized account.
Your integration checklist
- Capabilities
- What you need from your owner
- Create a Cloudflare account, choose the services and billing plans required by the workload, and authorize only the accounts and resources the agent should operate.
- Authentication
- Connect to https://mcp.cloudflare.com/mcp with OAuth or a scoped Cloudflare API bearer token. Account tokens need Account Resources Read for account discovery; tokens with client IP filtering are not supported by this MCP server.
- Your first task
- Check the connected identity and account, discover the read operation for an existing Worker, and inspect its configuration before making changes.
- Billing & payment
- Underlying products have separate plans and usage meters. Workers offers free allowances and paid usage; storage, databases and other attached services can add charges.
- Agent payment support
- Not verified. Programmatic usage does not establish autonomous checkout support.
- Operational limits
- API permissions and product runtime limits still apply. Broad MCP access can change live infrastructure. Large tool responses can be truncated, so bound queries and inspect truncation markers.
ORIGINAL AI AGENT STORE INTEGRATION PLAYBOOK
Resolve account and resource identity before executing a discovered operation
Editorial recommendation · Reviewed 2026-10-20
When to choose this service
Choose Cloudflare when the workflow needs managed compute, storage or networking that remains available independently of the agent process.
The scarce resource is operated remote infrastructure and network capacity. Writing a Worker or a DNS configuration does not supply the machines, storage and network that make it available.
Suggested workflow
- Record the intended account and resource IDs, then use read operations to establish the current state. Select the smallest credential scope that permits the approved task.
- Discover the exact operation and required fields. Prepare a bounded change in a development resource and check its runtime requirements and billing impact.
- Apply the approved change once, retain its resource or version identifier, and verify the resulting endpoint and configuration from a fresh read.
Acceptance checks before real work
- Match account identity before executing code that calls the API.
- Check for truncated lists before concluding a resource is missing.
- Verify the service endpoint independently of the management API response.
If the workflow fails
After a timeout, read the target state before retrying a mutation. Resolve missing permissions deliberately and restore the known configuration if the resulting service fails its health check.