Heroku MCP Server for Claude

Content

Heroku MCP Server connects Claude and other MCP-capable clients to the Heroku Platform through a local stdio server backed by the Heroku CLI. It gives an assistant tools for listing and inspecting apps, managing dynos, reading logs, provisioning add-ons, creating pipelines, promoting releases, toggling maintenance mode, deploying projects, running one-off dynos, operating Heroku Postgres, and optionally using Heroku AI tools when the Heroku AI CLI plugin is installed.

The safest setup is Heroku's recommended heroku mcp:start command, because it uses the active Heroku CLI session instead of storing a raw Heroku API token in MCP client configuration. Treat the connected client as an operator with the same Heroku access as the authenticated CLI session.

Features

• Official Heroku Platform MCP server maintained in the heroku/heroku-mcp-server repository.
• Stdio MCP server that runs through Heroku CLI 10.8.1 or later.
• Recommended heroku mcp:start launch path that reuses existing Heroku CLI authentication.
• Alternate npm launch path with npx -y @heroku/mcp-server and HEROKU_API_KEY.
• Client examples for Claude Desktop, Cursor, VS Code, Zed, Windsurf, Cline, and Trae.
• App management tools for listing apps, inspecting app details, creating apps, renaming apps, transferring apps, deploying projects, and running one-off dynos.
• Dyno and process tools for listing, scaling, resizing, and restarting dynos.
• Add-on tools for listing add-ons, inspecting add-ons, provisioning add-ons, and listing add-on services and plans.
• Maintenance and logs tools for enabling or disabling maintenance mode and reading application logs.
• Pipeline tools for creating, listing, inspecting, and promoting Heroku pipelines.
• Team and space tools for listing teams and Private Spaces.
• Heroku Postgres tools for SQL execution, database details, active queries, locks, outliers, credentials, process termination, maintenance, backups, and upgrades.
• Optional Heroku AI tools for listing available models, provisioning model access, and making inference requests when the @heroku/plugin-ai CLI plugin is installed.
• Dev Center resource support for bringing Heroku documentation into compatible MCP clients.

Use Cases

• Ask Claude to inventory Heroku apps, dynos, add-ons, teams, spaces, and pipelines before a migration or cleanup.
• Inspect app metadata and logs while debugging a failed deploy or runtime incident.
• Scale or restart dynos during a supervised staging incident-response drill.
• Create a disposable app or pipeline from an app.json project definition.
• Run a short-lived one-off dyno to test a script after reviewing the generated code and target app.
• Inspect Heroku Postgres locks, outliers, active queries, backup state, or maintenance windows.
• Execute SQL in development or staging databases while keeping production databases out of the MCP client's credentials.
• Pull Heroku Dev Center context into a Claude workflow while building or operating Heroku apps.

Installation

Preferred Heroku CLI setup

Install or upgrade the Heroku CLI, authenticate with the intended account, and confirm version 10.8.1 or later:

heroku --version
heroku auth:whoami

Then configure the MCP client to launch the server through the Heroku CLI:

{
  "mcpServers": {
    "heroku": {
      "command": "heroku mcp:start"
    }
  }
}

VS Code stdio setup

VS Code's MCP config can split the command and arguments:

{
  "mcp": {
    "servers": {
      "heroku": {
        "type": "stdio",
        "command": "heroku",
        "args": ["mcp:start"]
      }
    }
  }
}

npm setup with API token

Use the npm package only when you intentionally want to pass a Heroku authorization token through HEROKU_API_KEY:

{
  "mcpServers": {
    "heroku": {
      "command": "npx",
      "args": ["-y", "@heroku/mcp-server"],
      "env": {
        "HEROKU_API_KEY": "<YOUR_HEROKU_AUTH_TOKEN>"
      }
    }
  }
}

Generate a token with heroku authorizations:create, heroku auth:token, or the Heroku Dashboard's account applications page. Rotate and revoke tokens when they are no longer needed.

Configuration

Request timeout

The npm server supports MCP_SERVER_REQUEST_TIMEOUT, defaulting to 15000 milliseconds when unset:

{
  "mcpServers": {
    "heroku": {
      "command": "npx",
      "args": ["-y", "@heroku/mcp-server"],
      "env": {
        "HEROKU_API_KEY": "<YOUR_HEROKU_AUTH_TOKEN>",
        "MCP_SERVER_REQUEST_TIMEOUT": "30000"
      }
    }
  }
}

Use longer timeouts for operations that regularly wait on Heroku CLI output, but avoid masking hung or ambiguous deployment/database operations.

Heroku account scoping

Before connecting the server, run:

heroku auth:whoami
heroku apps --team YOUR_TEAM

Confirm the authenticated account, team, and target apps are the ones you want the assistant to see. If the workflow only needs staging access, authenticate a dedicated Heroku account or team role with staging-only permissions.

Examples

Ask Claude to inspect app and pipeline state:

Use the Heroku MCP server to list apps for the staging team, summarize dyno
formation and add-ons for APP_NAME, and do not change anything.

Ask for a supervised logs review:

Read the latest Heroku logs for APP_NAME, identify likely startup failures, and
propose fixes. Do not restart dynos or change config.

Run a database investigation only after scoping the target:

Use pg_info, pg_locks, and pg_outliers on the staging database for APP_NAME.
Do not run pg_psql, pg_kill, pg_credentials, pg_backups, or pg_upgrade unless I
approve the exact command and target.

Source Notes

• Official docs: Heroku MCP Server STDIO Mode.
• Official repository: heroku/heroku-mcp-server.
• npm package: @heroku/mcp-server, Apache-2.0, Node.js >=20.
• The Dev Center page says the server is early development and currently supports Heroku CLI 10.8.1 or later.
• The README and manifest list app, dyno, add-on, maintenance, log, pipeline, team, space, Postgres, deployment, one-off dyno, optional AI, and Dev Center resource capabilities.

Duplicate Check

• Checked current upstream content for Heroku titles, slugs, source URLs, heroku-mcp-server, @heroku/mcp-server, heroku mcp:start, and Dev Center MCP docs.
• Checked live open issues and pull requests for Heroku MCP and heroku-mcp-server.
• No dedicated Heroku MCP entry, source URL duplicate, target file duplicate, or open duplicate PR was found before drafting.

Disclosure

This is an independent community directory entry submitted by oktofeesh1. Heroku is a Salesforce platform, but this listing is not sponsored, paid, affiliate-backed, or submitted by Heroku. Use Heroku's current pricing, security, compliance, and support documentation when deciding whether to connect an AI assistant to real Heroku resources.