Honeybadger MCP Server

The Honeybadger MCP Server connects AI assistants to Honeybadger, a platform for error tracking and application monitoring. Error tracking means that errors from your running application are captured automatically, grouped and stored with their stack trace. MCP, the Model Context Protocol, is an open standard. Through it, an assistant such as Claude Code or Cursor calls external tools. The server comes from Honeybadger itself and is published on GitHub as an open source project under the MIT licence. It is built on the Honeybadger API. Instead of opening an error in the browser, copying the stack trace and explaining it to the assistant, the assistant fetches the details on its own.

Typical use: from error to fix

The documentation gives a simple workflow as an example. You hand the assistant the link to an error and ask it to fix it. The assistant then reads project and error details, opens the affected source file in your code and proposes a fix. You can also ask about recent error activity across your projects, filtered by time range and environment. What sets this apart from searching the dashboard is that the assistant ties error data directly to your source code. Honeybadger itself points out that you should review the results, because AI models make mistakes.

Hosted with OAuth or self-run

According to the changelog, a hosted server has been available since July 2026. It lives at https://mcp.honeybadger.io/mcp, and at https://eu-mcp.honeybadger.io/mcp for accounts in the EU. Each endpoint only accepts accounts from its own region. You need neither Docker nor a token in a configuration file. On first connection a browser window opens. There you pick an account and decide whether access is read-only or read and write. Tokens are short-lived and refresh in the background. Connections can be revoked at once under User Settings and API Access. Admins can review and revoke the connections of their team members as well.

Alternatively, you run the server yourself, as a Docker image from ghcr.io or as a Go binary you build. In that case it authenticates with a personal API token from your user settings. For EU accounts, set the HONEYBADGER_API_URL variable to the EU endpoint. According to the documentation, US and EU tokens are not interchangeable.

What can the Honeybadger MCP Server do?

The README lists tools across several areas. For projects, you can list, view, create and update projects and fetch occurrence counts. For faults, which is what Honeybadger calls grouped errors, there is search and filtering, individual notices with stack traces, affected users and comments. With query_insights the assistant runs queries in BadgerQL, Honeybadger's query language for event and log data. On top come tools for dashboards, alarms with their history, check-ins, uptime sites with outages, and deploys. A tool called search_tools helps the assistant find the right tool. The documentation for the hosted server still lists account and team management and uptime monitoring there as in development.

Safety: read-only by default

On the self-run server, read-only mode is on by default. Write tools such as create_project or delete_project are then missing from the list. Only HONEYBADGER_READ_ONLY=false unlocks them. Deletions also work in two steps: the first call only returns a preview and a confirmation token that is valid for ten minutes. Only the second call with that token actually deletes. With the hosted server you choose the access level when you sign in. According to the documentation, access never exceeds your own permissions.

Cost

The server itself is free and open source. You need a Honeybadger account. According to the pricing page, there is a free Developer plan for individual developers as well as paid Team and Business plans with more volume, longer retention and features such as SSO. You will find current terms on Honeybadger's official pricing page.

Who is the Honeybadger MCP Server for?

It pays off for teams that already use Honeybadger for error monitoring and work with an AI assistant in their editor. It is most useful in everyday debugging: the error, its stack trace and the source code end up in one conversation. If you need EU data residency, there is a dedicated EU endpoint. Start with read access and unlock write permissions only once you trust the workflow.

A Honeybadger account (for OAuth with the hosted server) or a personal API token (for running via Docker), and an MCP client with HTTP transport, for example Claude Code or Cursor.

With the Honeybadger MCP Server, an AI assistant reads errors, stack traces and Insights data from Honeybadger. This guide leads from choosing the region to the first error analysis.

Step 1: Prerequisites

You need a Honeybadger account with access to the project you want, and an MCP client with HTTP transport, such as Claude Code or Cursor. Check first whether your account lives in the US or the EU region. Each endpoint only accepts accounts from its own region.

Step 2: Connect it to Claude Code

For US accounts the command is claude mcp add --transport http honeybadger https://mcp.honeybadger.io/mcp. For EU accounts, register the address https://eu-mcp.honeybadger.io/mcp instead. In Cursor, Windsurf or Claude Desktop, add the same address as a URL to the MCP configuration. Then check with /mcp that the server is connected.

claude mcp add --transport http honeybadger https://mcp.honeybadger.io/mcp

Step 3: Sign in and choose the access level

On first use a browser window opens. Choose the account and the access level there. For getting started, read-only access is the safe choice. Tools for creating, updating and deleting are then missing from the list. Tokens are short-lived and refresh automatically. If your client keeps asking you to sign in again, it does not support refresh tokens, according to the documentation.

First error analysis

Copy the link to an error from Honeybadger and ask the assistant to fix it. It then reads the error details including the stack trace, opens the affected file and proposes a fix. Review the suggestion like any other code change. For an overview, ask for the most frequent errors in production over the past week.

Query Insights with BadgerQL

If you use Honeybadger Insights, the assistant can run queries in BadgerQL with query_insights. This is Honeybadger's query language for event and log data. That way you can analyse error trends over time without knowing the syntax yourself.

Step 4: Run it yourself with Docker

Without the hosted server, the image ghcr.io/honeybadger-io/honeybadger-mcp-server runs with a personal API token in HONEYBADGER_PERSONAL_AUTH_TOKEN. Here read-only mode is the default. Only HONEYBADGER_READ_ONLY=false unlocks write tools. For EU accounts, set HONEYBADGER_API_URL to https://eu-app.honeybadger.io and use an EU token.

Step 5: Revoke connections

You can see and revoke connected apps under User Settings and API Access. Account admins can revoke team members' connections under Account Settings and API Access. Revocation takes effect immediately.

OAuth with the Honeybadger account and short-lived tokens (hosted server) or a personal API token in HONEYBADGER_PERSONAL_AUTH_TOKEN (Docker mode).

Read: projects, errors with stack traces and affected users, Insights queries, dashboards, alarms, check-ins, uptime and deploys. Write (only when write access is granted): create, update and delete projects, dashboards, alarms, check-ins and comments.

Error data, stack traces, log events and user details that are read enter the AI client context and therefore reach the model provider. The hosted server is operated by Honeybadger in US and EU regions.

Error data can contain personal data and secrets, with write access projects and alarms can be deleted, API tokens sit in configuration files in Docker mode, and error messages can carry prompt injection.