Honeybadger MCP Server
Honeybadger MCP server: the agent reads errors and uptime checks, runs Insights queries and edits dashboards and alarms
- 25 ★ on the repository
- Last commit
- MIT
Extended access
Why this level
- A Honeybadger token opens error traces, comments and event environment data that can hold user data
- With `HONEYBADGER_READ_ONLY=false` the agent creates and deletes projects, dashboards, alarms and checks
Install
- Works in
Manual install
claude mcp add --transport http honeybadger https://mcp.honeybadger.io/mcpHosted server for the US region, sign in through OAuth. For the EU the address is `https://eu-mcp.honeybadger.io/mcp`.
Access and checks
How we reviewExtended accessThe tool runs code or makes network calls and only reads external data.
- A Honeybadger token opens error traces, comments and event environment data that can hold user data
- With `HONEYBADGER_READ_ONLY=false` the agent creates and deletes projects, dashboards, alarms and checks
This is third-party code. Review the repository files before installing.
What it does
The server connects the agent to the Honeybadger data API, an error and uptime monitoring service. The agent lists projects and faults, reads comments on them and looks at deploys and uptime checks. It writes Insights queries in BadgerQL, so event summaries are built straight from chat. Separate tools work with streams, dashboards, alarms and check-ins. The developer recommends the hosted server: `https://mcp.honeybadger.io/mcp` for the US region and `https://eu-mcp.honeybadger.io/mcp` for the EU one. Sign-in goes through OAuth, with no Docker image or personal token. The local setup with Docker or a Go build from source remains for clients that cannot reach remote servers. The `get_reference` tool gives the agent reference topics on BadgerQL, queries, charts, dashboards, alarms and the error model, and `search_tools` finds a tool by description. Deletion takes two calls: the first erases nothing and returns a preview with a confirmation token, and the second with the same token performs the delete.
Who it is for. Developers and on-call engineers who watch errors in Honeybadger and want to triage them, build Insights queries and set up alarms from Claude Code, Cursor or VS Code.
Good fit when
- After a release the error count grew, and the agent compares faults by deploy, reads comments and names the suspect commit
- A weekly event summary needs an Insights query and you rarely write BadgerQL
- An error-rate alarm is created or edited from chat together with the dashboard it belongs to
- Check-ins and uptime checks are spread across projects and the agent lists the ones failing now
Not a fit when
- Errors are collected by another service: the server talks to Honeybadger only
- The agent must not touch project records: with `HONEYBADGER_READ_ONLY=false` it creates and deletes projects, dashboards and alarms
- Your Honeybadger account sits in a different region than the token: US and EU tokens do not work across regions
Example request
Show the 10 most frequent errors of the project over the last day and compare them with the previous deployMCP
- Transport
- http, stdio
- Authentication
- OAuth
Environment variables
HONEYBADGER_PERSONAL_AUTH_TOKENrequired, secret- API Token (`hbu_` or `hba_`) for local runs
HONEYBADGER_READ_ONLY- Defaults to `true`: no write tools. `false` enables create and delete
HONEYBADGER_API_URL- API base URL; `https://eu-app.honeybadger.io` for the EU region
Limitations
- Local mode needs Docker and an API Token with the `hbu_` (personal) or `hba_` (account) prefix. The old personal auth token is no longer accepted.
- The EU region needs a token from EU settings and `HONEYBADGER_API_URL=https://eu-app.honeybadger.io`. A US token does not work there, and the other way round.
- The local server starts in read-only mode. `HONEYBADGER_READ_ONLY=false` enables writes.
- The `get_project_report` tool is gone in the API v3 version: `query_insights` answers the same questions.
How to disable. The `claude mcp remove honeybadger` command removes the server from Claude Code; in other clients delete the `honeybadger` entry from the MCP configuration. A personal token is revoked on the Authentication tab in Honeybadger user settings.
FAQ
How do I connect Honeybadger to Claude Code?
Run `claude mcp add --transport http honeybadger https://mcp.honeybadger.io/mcp` and sign in through OAuth. For the EU region replace the address with `https://eu-mcp.honeybadger.io/mcp`.
How does the hosted server differ from the local one?
The hosted server signs in through OAuth and needs no Docker and no token in configuration. The local one runs from the `ghcr.io/honeybadger-io/honeybadger-mcp-server` image with your own API Token and is for clients without remote server support or for development.
Can the agent delete a project or an alarm?
The local server is read-only by default. With `HONEYBADGER_READ_ONLY=false` creating and deleting appear. Delete tools take two calls, where the first only shows a preview and issues a `confirm` token.
Which token does the local server need?
An API Token starting with `hbu_` (personal, created on the Authentication tab of user settings) or `hba_` (account). It goes in the `HONEYBADGER_PERSONAL_AUTH_TOKEN` variable; the name is kept from older versions.
What should I do after upgrading from an earlier version?
Replace the old personal token with an API Token. The `get_project_report` tool is gone; use `query_insights` instead, since every error occurrence is a `notice` event in the stream and the `ts` parameter sets the time window.
Discussion
No questions or reviews yet.
Related
MCP serversDatadog MCP Server (winor30)
Datadog MCP Server
Independent Datadog MCP server: an agent reads incidents, monitors, logs, metrics, traces and RUM and schedules downtimes.
A Node.js manager for MCP servers: it starts local servers, connects remote ones and gives a client one address, with a REST API for control
The built-in OneUptime MCP: an agent works incidents, reads logs and traces, acknowledges alerts and sets up monitors.
Datadog's official managed MCP server: logs, metrics, traces and incidents right in your agent
maintenant monitoring MCP server: an agent reads container logs, check history and alerts but never touches the containers
A self-hosted gateway that merges several MCP servers into one endpoint with auth and tool selection
README in short
The page opens by recommending the hosted server and then describes local setup for Cursor, Windsurf, Claude Desktop, Claude Code, VS Code and Zed. It continues with environment variables, EU region setup, command-line flags and the tool list grouped by area with parameters.