Heroku MCP Server
Connect Heroku to your AI agents with the Heroku MCP server.
Updated
What is Heroku MCP?
The Heroku MCP server is a Model Context Protocol server that connects Heroku to AI agents and assistants as a set of tools. With it connected, an agent can work with Heroku (paas hosting — apps, dynos, add-ons, config vars, releases, logs) directly — instead of you copying data in and out by hand.
It is a remote MCP server that uses OAuth, so you point your client at the server URL, authorize once, and the Heroku tools become available to the model.
Connect the Heroku MCP server
Connect Heroku MCP to Claude Code
- 1
Add the server
Use `claude mcp add` with the Heroku MCP server URL (or its `npx` command), following the official docs.
- 2
Authorize
Complete authentication (OAuth) so Claude Code can call the server.
- 3
Verify
Ask Claude to use a Heroku tool to confirm the connection is live.
Connect Heroku MCP to Cursor
- 1
Open MCP settings
In Cursor, go to Settings → MCP → Add new server.
- 2
Add Heroku
Provide the Heroku MCP server URL or command and complete authentication.
- 3
Test
Reference Heroku from chat to confirm Cursor can reach the tools.
Using Heroku MCP with Gamut
On Gamut, the Heroku MCP server becomes a tool your agents use inside an automated, event-driven workflow — triggered by an event or a schedule, acting through Heroku and your other connected MCP servers, with the connection managed by Gamut rather than a local config file.
Frequently asked questions
What is the Heroku MCP server?
It is a Model Context Protocol server that exposes Heroku to AI agents as tools so they can work with it (paas hosting — apps, dynos, add-ons, config vars, releases, logs).
How do I connect the Heroku MCP server?
Point your MCP client (Claude, Cursor, or Gamut) at the Heroku MCP server URL and complete authentication (OAuth). See the official documentation for the exact URL and scopes.
Can I use Heroku MCP with Gamut?
Yes — Gamut supports the Heroku MCP server, so your agents can use it as a tool in automated, triggered workflows.