Heroku AI: Heroku Remote MCP Server

Today, we’re thrilled to announce a new way in which agents can access the Heroku platform using the Heroku Remote MCP Server, now available at https://mcp.heroku.com/mcp.

This new remote server is an expansion of our earlier stdio-based MCP server and comes with secure OAuth authentication. It’s designed to provide a secure, scalable, and incredibly simple way for agents to interact with the Heroku platform and use tools to perform actions such as creating a Heroku app from your favorite agents such as Claude, Agentforce, or Cursor. With Agentforce 3.0 announcing native support for MCP, you can bring Heroku Remote MCP Server to Agentforce.

Secure access and authentication with remote MCP servers

If you’re new to MCP, read this introduction to MCP to familiarize yourself. While our initial stdio MCP server supports local development capabilities by allowing agents to interact with the Heroku platform as a subprocess, it was limited to tethering your agent’s capabilities to a single machine. The new Heroku Remote MCP Server overcomes this with enhanced security for your AI workflows by centralizing access to the Heroku platform. The Heroku Remote MCP Server is easily accessible through clients that support remote servers and uses the industry-standard OAuth 2.0 protocol. When you connect a new client, you’ll be prompted to authenticate with your Heroku account, giving you clear and user-consented control over which tools can be accessed by your client.

How to connect your MCP client

As long as your agent supports remote MCP servers with OAuth, you can connect to Heroku in a few easy steps.

1. Configure the server: Following your client’s documentation, add a new MCP server and use the URL: https://mcp.heroku.com/mcp
2. Authenticate: Your client will redirect you to login with the Heroku account via OAuth.
3. Connect: Once you authenticate, your client is securely connected and ready to go!

Claude Desktop (Free)

For the Claude desktop application, you can connect using a proxy command.

1. Open the configuration file located at ~/Library/Application Support/Claude/claude_desktop_config.json.
2. Add the following entry to the mcpServers object and restart the Claude app:

{
    "mcpServers": {
      "Heroku": {
        "command": "npx",
        "args": ["-y", "mcp-remote", "https://mcp.heroku.com/mcp"]
      }
    }
}

Cursor

For Cursor, you can connect on the Tools & Integrations section in the Cursor settings page.

1. Click on Add a Custom MCP.
2. Add the following entry to the mcpServers object.
3. Go back to Tools & Integrations, click on “Needs login” to the right of the Heroku MCP entry and authenticate via OAuth.

{
    "mcpServers": {
      "heroku": {
        "url": "https://mcp.heroku.com/mcp"
      }
    }
}

VSCode

For Visual Studio Code, open the command palette and select the following:

1. Select MCP: Add Server....
2. HTTP (HTTP or Server Sent Events)
3. Insert the url https://mcp.heroku.com/mcp.
4. Insert heroku as server id.

This will add the MCP configuration to the settings.json file, from here it will ask to start the OAuth authentication.

Extend your agents with the Heroku Remote MCP Server

The Heroku Remote MCP Server empowers your agent with a rich set of tools to understand and interact with the Heroku platform. Your agent can now perform a wide array of tasks on your behalf:

• Manage application lifecycle: list your apps, get info on a specific app, and even create and update applications directly via natural language.
• Info: Get a list of the teams you belong to, and view your Private Spaces.
• Add-on marketplace: List available Heroku Add-ons and check their various plans.

This is just the initial set of tools that we have enabled. We are continuously working to enable additional tools to help you with an increasing variety of workflows.

What’s next

We’re excited about the rapid innovation in the MCP and AI ecosystem and are keeping close to the community. We expect to make updates to our MCP tools as the protocol evolves and from customer feedback. We at Heroku are obsessed with providing the best developer and operator experience for your AI workflows and agents. We started this journey with the launch of Heroku Managed Inference and Agents and support to build stdio MCP servers on Heroku and the Heroku Remote MCP Server (mcp.heroku.com/mcp) is the next exciting milestone on this journey.

Join the official Heroku AI Trailblazer Community to keep up with the latest news, ask questions, or meet the team.

To learn more about Heroku AI, check out our Dev Center docs and try it out for yourself.