Expose your MCP endpoint

What your endpoint is

Your MCP endpoint is your company, exposed as a hosted MCP server. It presents your unified data and your connected tools to any external AI agent over the standard Model Context Protocol — no per-client integration, no scraping, no fragile glue. One URL, one key, every agent.

The endpoint lives at:

https://businessmcp.com/api/mcp/mcp

It speaks the streamable-HTTP transport and is authenticated with a Bearer mcph_* key. There is nothing to deploy — it is provisioned the moment your workspace exists.

Create an mcph_* key

In your dashboard, open Settings and create an API key. Keys are prefixed mcph_ (a stable contract — do not expect that to change) and are shown once, so copy it somewhere safe. The key encodes which workspace the caller acts as; anyone holding it can call your endpoint as your workspace, so guard it like a database password.

Issue a separate key per agent or integration. That way you can revoke one without disrupting the others, and your audit log tells you exactly which caller did what.

Call the endpoint

MCP over HTTP uses JSON-RPC. Two calls cover most of what you need: discover the tools, then invoke one.

List available tools:

curl -X POST https://businessmcp.com/api/mcp/mcp \
  -H "Authorization: Bearer mcph_your_key_here" \
  -H "Content-Type: application/json" \
  -H "Accept: application/json, text/event-stream" \
  -d '{"jsonrpc":"2.0","id":1,"method":"tools/list"}'

Call a tool — here, pull analytics for the last 30 days:

curl -X POST https://businessmcp.com/api/mcp/mcp \
  -H "Authorization: Bearer mcph_your_key_here" \
  -H "Content-Type: application/json" \
  -H "Accept: application/json, text/event-stream" \
  -d '{
    "jsonrpc": "2.0",
    "id": 2,
    "method": "tools/call",
    "params": {
      "name": "get_analytics",
      "arguments": { "range": "30d" }
    }
  }'

The server authenticates the key, resolves your workspace, runs the tool against your unified data, and returns a structured result. Most real usage happens through an MCP client rather than raw curl — see connect Claude, GPT & Gemini.

What tools are exposed

Your endpoint surfaces two families of tools. Platform tools are built in: tasks, notes, business context, analytics (get_analytics), CRM lookups, goals, funnels and campaigns, plus action tools like send_email (approval-gated). Connector tools come from whatever you have linked in Connections — namespaced like github_*, gsc_*, slack_* — so connecting a tool once makes it available to every agent through this single endpoint.

Which tools a given key can reach is governed per connection, and the total is capped to keep model accuracy high. Connect more, and your endpoint grows more capable without you touching any protocol code.

Keeping it secure

The endpoint enforces the practices from our best practices guide for you: bearer authentication, per-connection tool allowlists, a per-run cost ceiling, event-based quotas, approval gates on sensitive actions, and a full audit log of every call. Your job is smaller but essential — rotate keys you no longer use, scope keys to the smallest surface each agent needs, and never commit a mcph_* key to source control.