March 20, 2026·4 min read

Why I Built gitglimpse

Every developer has the same standup problem. The answer was in the git log all along.

open-source developer-tools python

The standup problem every developer faces

It's 9:01 AM. Standup starts in sixty seconds. You stare at Slack, trying to reconstruct what you actually did yesterday. You remember fixing that auth bug, but what else? There were a few PRs, some review comments, maybe a config change. The details blur together. So you mumble something vague and move on.

The git log already has the answer

Your commit history is a near-perfect record of your workday. Every feature, every fix, every refactor — timestamped and annotated. The problem isn't missing data. It's that git log dumps a wall of hashes and messages that no one wants to parse at 9 AM. gitglimpse reads that history and turns it into a human-readable summary you can paste straight into Slack.

text

$ glimpse standup
Author: [email protected] · Context: commits

Standup — March 20, 2026

Yesterday:
  • Fixed token refresh race condition in auth middleware (feat/auth, ~0.5h)
  • Added retry logic to webhook delivery pipeline (main, ~1.5h)
  • Updated API rate-limit docs for v2 endpoints (main, ~0.5h)

Estimated effort: ~2.5h

Why existing tools miss the mark

Tools like GitHub Insights, LinearB, and Waydev are built for engineering managers tracking team-wide metrics. They require OAuth integrations, cloud dashboards, and seat-based pricing. If you just want to know what you did today — as a developer, for yourself — they are wildly over-scoped. gitglimpse is the opposite: a single CLI command, zero accounts, zero network calls.

What makes gitglimpse different

•Fully offline — reads only your local .git directory, never phones home.
•Effort estimation built in — approximates how long each task took based on commit timestamps and diff size.
•Noise filtering — merge commits, lock files, and formatting changes are automatically excluded.
•Ticket detection — branch names like feature/PROJ-123 are parsed automatically.
•Multi-project support — run from a parent directory to aggregate work across repos.
•LLM-optional — works perfectly without AI. Add --local-llm for richer summaries that read your actual code diffs, or pipe --json into Claude Code or Cursor.
•JSON mode for automation — structured output means you can build workflows, dashboards, or Slack bots on top of it.