Heroku MCP Server – Apps and Platform Operations
Heroku's official MCP servers give AI agents structured access to Heroku Platform resources. Use them to inspect, manage, deploy, scale, monitor, and troubleshoot Heroku applications and related services from an MCP-compatible client.
Overview
Heroku's official MCP offering connects compatible AI assistants to the Heroku
Platform. Heroku provides both a managed remote MCP server and a local STDIO
server, allowing agents to work with live Heroku applications and resources
through natural-language workflows instead of relying only on manually copied
CLI output or dashboard information.
What the MCP server enables
The Heroku MCP server is designed to let agents read, manage, and operate
Heroku Platform resources. Depending on the authenticated account and client
capabilities, an AI agent can support workflows such as:
- Inspecting Heroku apps and platform resources.
- Deploying, restarting, scaling, and monitoring applications.
- Viewing logs and troubleshooting application behavior.
- Working with dynos and process-related operations.
- Discovering and managing add-ons.
- Supporting Heroku Postgres and other database operations.
- Performing multi-step app lifecycle and performance workflows.
- Operating across Common Runtime, Cedar Private and Shield Spaces, and Fir
Private Spaces where supported by the account.
When to use it
Use Heroku MCP when an AI workflow needs live Heroku context or must perform an
approved platform operation. Typical examples include checking why an app is
failing, reviewing logs, scaling dynos, restarting an app, inspecting add-ons,
preparing deployment steps, troubleshooting a Heroku Postgres issue, or helping
an agent manage routine app lifecycle tasks from an IDE.
Connection and authentication
The hosted remote endpoint is https://mcp.heroku.com/mcp. It uses web-based
OAuth 2.0 through Heroku's identity provider and requires a client that supports
Streamable HTTP and OAuth token refresh.
The local STDIO server can be launched with heroku mcp:start, available in
Heroku CLI version 10.8.1 and later. Heroku recommends this method because it
reuses the current Heroku CLI authentication context and avoids exposing an API
key. A secondary local option is npx -y @heroku/mcp-server, which requires
HEROKU_API_KEY.
Key considerations
The STDIO server is described as early development and tools may evolve. The
remote server offers a centrally managed OAuth-based alternative to local
STDIO, while the local server can be useful when a client expects a local
process. Tool calls can affect real Heroku applications, deployments, scaling,
add-ons, databases, and logs, so require human confirmation for write or
production-impacting actions. Use least privilege, protect API keys, keep the
Heroku CLI updated, and reset OAuth tokens if a client connection becomes
invalid or hangs.
Supported Transports
streamable_http
URL: https://mcp.heroku.com/mcp
stdio
Command: heroku
Args:
mcp:start
stdio
Command: npx
Args:
-y@heroku/mcp-server
Frequently Asked Questions
- When should an AI agent use the Heroku MCP server?
- Use it when a workflow needs live Heroku application or platform context, such as inspecting apps, viewing logs, deploying, restarting, scaling dynos, troubleshooting failures, managing add-ons, or supporting Heroku Postgres and app lifecycle operations.
- What does the Heroku MCP server add to an AI agent's capabilities?
- It gives the agent structured access to current Heroku Platform resources and supported operations, allowing it to work with real apps, dynos, logs, add-ons, deployments, and databases instead of relying only on static model knowledge or manually copied CLI output.
- What can an AI agent access or manage through Heroku MCP?
- Depending on account permissions and the selected transport, the agent can work with Heroku apps, dynos, logs, deployments, scaling, monitoring, add-ons, database-related operations, and supported platform resources across Heroku runtimes and spaces.
- How is authentication configured for the Heroku MCP server?
- The hosted remote server uses web-based OAuth 2.0 through Heroku's identity provider. The recommended local server uses the current Heroku CLI session through `heroku mcp:start`. The npx local option requires HEROKU_API_KEY, which must be stored securely outside source control.
- Which transport should be used for the Heroku MCP server?
- Use Streamable HTTP with https://mcp.heroku.com/mcp for Heroku's managed remote service. Use stdio with `heroku mcp:start` for local client integrations that can reuse the Heroku CLI session, or the npx stdio option only when an API-key-based local setup is required.