Google's open-source MCP server for enterprise databases — one binary that fronts AlloyDB, BigQuery, Spanner, Postgres, MySQL, Oracle, MongoDB, Snowflake and more.
Databases & DataCloud & Infrastructure
honeycombobservabilitytracingopentelemetrysloincident-responsebubbleup
Summary
Honeycomb's hosted MCP server — query traces, metrics, logs and events, run BubbleUp, and read Boards, Triggers and SLOs from an AI agent.
Features
- Hosted by Honeycomb at mcp.honeycomb.io — nothing to install
- US and EU regional endpoints
- Query traces, metrics, logs and events; inspect whole traces compactly
- BubbleUp analysis, service maps and regex column search
- Read Boards, Triggers, SLOs and notification recipients
- OAuth 2.1 for interactive clients; Management API key for headless agents
Installation
Set up this MCP server in your favorite AI agent — copy a ready-made configuration below.
Description
Honeycomb's MCP server gives an agent the thing it usually lacks during an incident: the ability to ask production a question and get a real answer back. It is a Honeycomb-hosted endpoint, so there is no server to run — you point an MCP client at https://mcp.honeycomb.io/mcp (or https://mcp.eu1.honeycomb.io/mcp for EU environments) and authorize in a browser.
What the agent can actually do
The tool surface tracks how engineers use Honeycomb rather than wrapping the REST API one-to-one. Agents can discover the workspace and its columns — including regex column search, which matters on wide datasets where finding the right field is most of the work — then run queries across traces, metrics, logs and events. Trace and span inspection returns a whole trace in compact form, which is what makes an agent capable of spotting an n+1 query or a critical path on its own rather than describing one in the abstract.
Beyond querying, there is BubbleUp analysis for explaining why one slice of traffic differs from the rest, service maps, semantic-convention lookups, and read access to Boards, Triggers, SLOs and notification recipients — so "is this already alerting, and is the SLO burning?" is answerable in the same turn as the query that raised it. Canvas investigations and AI conversation analysis are exposed too, along with a feedback tool the agent can call directly.
Two ways to authenticate
Interactive use runs on OAuth 2.1: the client opens a browser once and reuses the grant afterwards. For headless agents and CI, generate a Honeycomb Management API key with Model Context Protocol and Environments read permissions, and pass it as a bearer token through mcp-remote.
Setup
Claude Code takes one command: claude mcp add honeycomb --transport http https://mcp.honeycomb.io/mcp. Cursor and VS Code have one-click installers, Claude Desktop connects it as a custom connector, and Amazon Q Developer CLI bridges via npx mcp-remote.
A note on the older server
Honeycomb previously shipped a self-hosted server at honeycombio/honeycomb-mcp, which required Node.js, an Enterprise plan and local STDIO configuration. That repository is now deprecated in favour of the hosted endpoint — new setups should use the URL above rather than building from source.
Related MCP Servers
Datadog's official managed MCP server — query logs, metrics, traces, incidents and dashboards from an AI agent, inside your existing Datadog permissions.
Cloud & Infrastructure
Official Grafana MCP server: search and edit dashboards, query Prometheus, Loki and Pyroscope, and work with alerts, incidents and on-call from an agent.
Cloud & Infrastructure
1 views
Hostinger's official MCP server, shipping 278 tools across VPS, hosting, domains, DNS, mail, WordPress and billing — split into per-domain binaries you can scope down.
Cloud & Infrastructure
1 views