API Connect MCP server

<!-- mcp-name: io.github.bvenkata/mcp-api-connect -->

mcp-api-connect&trade;

One payload in, any API out. mcp-api-connect is a protocol- and auth-agnostic connector engine: describe a target service (URL, protocol, auth, request/ response shape) once, then send it a normalized payload and get a normalized response back — whether the target is a REST/JSON API, a legacy SOAP service, protected by an API key, Basic auth, a Bearer token, or OAuth2 client credentials.

It ships as three things built on the same core engine, so however you want to use it, you can:

directly, no server required.

• A Python library — pip install mcp-api-connect, call MCPAPIConnectEngine

mcp-api-connect-api, POST to /invoke.

• A standalone HTTP API — pip install mcp-api-connect[api], run

point any MCP client (Claude, etc.) at it so an agent can call registered connectors — or arbitrary services on the fly — as tools.

• An MCP server — pip install mcp-api-connect[mcp], run mcp-api-connect,

Why

Every integration project reinvents the same wheel: a REST client here, a SOAP client there, one auth flow per service, ad-hoc request/response mapping scattered across the codebase. mcp-api-connect centralizes that into one declarative spec (InvokeSpec) and one execution engine, so adding a new target service is config, not code.

Quick start (library)

pip install mcp-api-connect

import asyncio
from mcp_api_connect import MCPAPIConnectEngine, InvokeSpec, Target, AuthSpec, AuthType, RequestFormat, ResponseFormat

spec = InvokeSpec(
    target=Target(base_url="https://api.example.com"),
    auth=AuthSpec(type=AuthType.API_KEY, config={"api_key": "secret", "header_name": "X-API-Key"}),
    request_format=RequestFormat(method="POST", path="/v1/orders", content_type="json"),
    response_format=ResponseFormat(content_type="json"),
)

async def main():
    async with MCPAPIConnectEngine() as engine:
        result = await engine.invoke(spec, {"customer": "jane"})
        print(result.success, result.data)

asyncio.run(main())

Quick start (HTTP API)

pip install "mcp-api-connect[api]"
mcp-api-connect-api   # serves on :8000, interactive docs at /docs

curl -X POST http://localhost:8000/invoke -H 'content-type: application/json' -d '{
  "spec": {
    "target": {"base_url": "https://api.example.com"},
    "auth": {"type": "api_key", "config": {"api_key": "secret"}},
    "request_format": {"method": "POST", "path": "/v1/orders"},
    "response_format": {"content_type": "json"}
  },
  "payload": {"customer": "jane"}
}'

Register a reusable connector once, then invoke it by name:

curl -X POST http://localhost:8000/connectors -d '{"name": "orders-api", "spec": {...}}'
curl -X POST http://localhost:8000/connectors/orders-api/invoke -d '{"customer": "jane"}'

Quick start (MCP)

pip install "mcp-api-connect[mcp]"

{
  "mcpServers": {
    "mcp-api-connect": { "command": "/path/to/.venv/bin/mcp-api-connect" }
  }
}

Or run it in a container (stdio transport):

docker build -t mcp-api-connect .
docker run --rm -i mcp-api-connect

Exposes tools: invoke (stateless, one-off), registerconnector, listconnectors, invokeconnector (by name), deleteconnector. An agent can register a connector for "the Salesforce API" once, then just say "call it with this payload" from then on.

➜ Full setup for Claude Desktop / Claude Code / Cursor, persistence, security notes, and a worked example: docs/mcp-integration.md.

Core concepts

Shortened. The full README is on GitHub.