Sentry — post-mortems with real data

New errors since the deploy, issue deep-dives, code-to-error links — inside the post-mortem note.

MCP · Sentry — post-mortems and error triage

Use case: "what broke last night?" — production errors analyzed inside the post-mortem note, with stack traces and frequencies.

Configuration (.znote/mcp.json)

{
  "mcpServers": {
    "sentry": {
      "command": "npx",
      "args": ["-y", "mcp-remote", "https://mcp.sentry.dev/mcp"]
    }
  }
}

Official remote server (OAuth on first run). Self-hosted: the @sentry/mcp-server package over stdio with SENTRY_ACCESS_TOKEN and SENTRY_HOST in env.

Exposed tools: find_issues, get_issue_details, find_errors_in_file, update_issue, find_releases, get_trace_details…

Examples

The morning after a deploy

Which new errors appeared in the last 24h on the "web-app" project?
For the top 3: message, frequency, users affected, and your reading of
the probable cause from the stack trace.

Structured post-mortem

Detail issue WEB-APP-1234: first/last occurrence, volume, breadcrumbs.
Draft the start of a post-mortem: impact, cause hypothesis, proposed
actions.

Code ↔ errors link

Are there recent errors pointing at the file src/api/checkout.js?
If so, which ones and at which lines?

Tips

Combine with GitHub: "for each top-3 error, is there an open GitHub issue? If not, create it with the stack trace".
The post-mortem note keeps the results: the incident's context is still readable six months later.