Summary
Incident debugging in Sentry, with three tools instead of thirty.
It covers the path you actually walk during an incident — open the issue, find the event that matters, follow the trace — and stops there. Being one binary means nothing to install alongside it and a much smaller tool list riding in context.
What it is
Three focused Sentry tools — issue detail with stacktraces and tags, event search inside an issue, and trace detail with the span tree — shipped as one Rust binary.
What you get
- Full issue detail: metadata, tags, stacktraces, and optionally one specific event
- Search within an issue's events using Sentry's own query syntax
- Trace detail with the span tree and timing, for distributed tracing
- A single binary with no Node or Python runtime behind it, and an instant cold start
- A deliberately short tool list, which keeps the tool definitions small in every turn
- SOCKS5 and HTTPS proxy support
Requirements
SENTRY_AUTH_TOKEN and SENTRY_HOST — sentry.io or your own instance hostname. Install with cargo install sentry-mcp, from git, or with mise; the binary is sentry-mcp. Standard proxy environment variables are picked up if you need to route through one.
Setup effort
One command plus a key — cargo install sentry-mcp, then supply credentials