# Connect AI assistants (MCP)

> Paste https://mcp.heyrik.com into Claude, ChatGPT, Cursor or any MCP client, sign in, and the assistant can work with your whole HeyRik account — placing calls only if you allow it.

Source: https://www.heyrik.com/docs/guides/connect-ai-assistants

Paste https://mcp.heyrik.com into Claude, ChatGPT, Cursor or any MCP client, sign in, and the assistant can work with your whole HeyRik account — placing calls only if you allow it.

4 minute read Dashboard guide

## The HeyRik MCP server

HeyRik runs a remote Model Context Protocol server at https://mcp.heyrik.com. Paste that URL into your AI app, sign in to HeyRik in the browser window that opens, and approve. Every tool calls the matching public API endpoint as your account, with the same permissions and data scoping.

01

### URL

https://mcp.heyrik.com (Streamable HTTP; https://mcp.heyrik.com/mcp works too).

02

### What it can do

Your whole account: agents (create, change, delete), knowledge, voices, phone numbers, calls and transcripts, campaigns, CRM leads, widgets, chat history, usage and analytics.

03

### API key

Send Authorization: Bearer YOUR\_API\_KEY. Works with Claude Code, Claude Desktop and Cursor. A key has every permission.

04

### Sign in (OAuth)

claude.ai, ChatGPT, Claude Desktop and Claude Code can connect without a key: you sign in to HeyRik and approve the app.

## Connect your assistant

Use the HeyRik endpoint URL in your assistant's connector settings.

01

### Claude Code

claude mcp add --transport http heyrik https://mcp.heyrik.com — then run /mcp → heyrik → Authenticate (or add --header "Authorization: Bearer YOUR\_API\_KEY").

02

### Claude Desktop and claude.ai

Settings → Connectors → Add custom connector → name HeyRik, URL https://mcp.heyrik.com → Connect, then sign in and approve on the HeyRik page that opens.

03

### ChatGPT

Settings → Apps & Connectors → Advanced → enable Developer mode, then Create a connector with URL https://mcp.heyrik.com and OAuth authentication. Sign in to HeyRik and approve.

04

### Cursor

In ~/.cursor/mcp.json: { "mcpServers": { "heyrik": { "url": "https://mcp.heyrik.com", "headers": { "Authorization": "Bearer YOUR\_API\_KEY" } } } }

## Choose what the assistant may do

The approval page shows the app's name, where you return to, and the permissions it asks for.

01

### Read

Agents, voices, phone numbers, knowledge, calls (transcripts, summaries, extracted data, recording links), campaigns and credit balance.

02

### Write

Create, change and delete agents, knowledge, campaigns (never start them), CRM leads and widgets, and route phone numbers. Deleting always asks you first.

03

### Place calls

Place real calls, start campaigns, add campaign contacts and call CRM leads. Off by default — tick it only if you want the assistant to dial.

## Costs and safety

Read tools cost nothing. Dialling is always explicit.

01

### Real calls

place\_call, start\_campaign, add\_campaign\_contacts and call\_leads dial real numbers and spend credits (about 1 credit per call minute). They run only with confirm: true and the Place calls permission.

02

### Limits

No MCP request limit per connection; lists return up to 100 items per call (page through for more).

03

### Secrets

Campaign ingest tokens and similar secrets are never returned to the assistant.

## Disconnect an app

Every approved app has its own key you can revoke.

01

### Revoke

Each approved app appears on the API keys page as “MCP: <app name>”. Revoke it there and the app loses access immediately.

02

### Reconnect

Approving the same app again replaces its previous connection.