MCP server

LayerCall speaks the Model Context Protocol, so an AI agent can check a phone number, email, IP or domain in the middle of a conversation instead of you writing HTTP calls around it. MCP is an open protocol — this works identically in Claude, Cursor, ChatGPT, VS Code, Windsurf, Zed and anything else that speaks it. The only thing that differs between them is which file the config goes in.

Server URL

https://www.layercall.com/api/mcp

Transport is Streamable HTTP. Authenticate with your API key as a bearer token — a free key covers 1,000 lookups a month.

Setup

Pick your client.

Run this in your terminal — it writes the config for you.

claude mcp add --transport http layercall https://www.layercall.com/api/mcp \
  --header "Authorization: Bearer YOUR_KEY"

Replace YOUR_KEY with a LayerCall API key. A test-mode key works here too and never bills — the right choice while an agent is calling things in a loop.

Tools

Five tools, each returning the same shape as its REST endpoint.

Every tool takes an optional strictnessof 0–3 and returns a 0–100 risk_score with an allow / review / block verdict.

Billing

A tool call costs exactly what the equivalent REST call costs and draws on the same monthly quota — the connector is not a way around it. Test-mode keys work here and never bill, which is what you want while an agent is calling things in a loop.

Checking it works

initialize and tools/list answer without a key, so you can confirm the endpoint is reachable before configuring credentials.

curl -X POST https://www.layercall.com/api/mcp \
  -H "Content-Type: application/json" \
  -d '{"jsonrpc":"2.0","id":1,"method":"tools/list"}'