Engineering routines

The maintenance work nobody schedules, scheduled.

Engineering routines take the recurring upkeep that competes with feature work — triaging new issues, chasing stalled pull requests, reading error trends — and put it on a clock. Each run reads the connected tools, proposes concrete next actions with evidence, and only acts through tools you have explicitly approved.

Routines in engineering

• Every weekday

  GitHub issue triage

  Keep the issue tracker answerable by triaging every new report before a human looks at the queue.

• Weekly

  Stale pull request review

  Stop finished work from dying in review by naming what blocks each stalled pull request.

• Weekly

  Weekly error trend review

  Notice which errors are actually trending before users report them, without staring at dashboards.

Works with

• L

  Linear

  Turn recurring findings into triaged issues, project updates, and accountable follow-up work.

• Se

  Sentry

  Review incidents and error trends on a schedule, then surface the changes that need attention.

• GH

  GitHub

  Read repositories and issues, then open reviewable pull requests through GitHub's remote MCP server.

See every integration

Common questions

Can a routine modify our repository?

Only through tools you explicitly enable, and write-like actions need a separate automatic-execution approval. A triage routine can run for months in propose-only mode.

What happens when there is nothing to report?

The run says so explicitly and finishes. A quiet week produces a short all-clear rather than padding, and the trace shows what was checked.

Which tools do engineering routines use?

GitHub is built in, and Linear and Sentry connect over MCP. Each routine lists the capabilities it needs, and you grant them per connection.