Model Context Protocol on Heroku

Host MCP servers, wire them to agents, and extend your AI stack with Heroku infrastructure.

The Model Context Protocol (MCP) lets large language models call external tools, access resources, and render interactive widgets. Heroku provides everything you need to host MCP servers, expose manifests, and connect them to popular clients such as ChatGPT, Claude, Cursor, and Agentforce.

Use this section as the control center for MCP:

• Server modes – Deploy STDIO or Remote MCP servers on Heroku with built-in SSL, scaling, and logging.
• Tools – Combine Heroku tools (CLI, PG, KV) with your own MCP endpoints for richer agents.
• Cookbooks – Launch ChatGPT apps, integrate reference implementations, and follow end-to-end deployment guides.
• Clients – Configure each agent platform to consume your Heroku-hosted manifest.

Quick tour

Task	Where to start
Deploy an MCP server	STDIO mode or Remote mode
Connect to ChatGPT, Claude, Cursor, Agentforce	Client guides
Build rich ChatGPT apps with widgets	ChatGPT apps cookbook
Combine tools and automate flows	Heroku tools reference

Bring your existing APIs, data stores, or automation scripts. MCP simply packages them so agents can discover and invoke the right capability at the right time. Heroku takes care of the runtime so you can focus on the tools themselves.