jzb1006

A vigil is a watch kept while others rest. Vigil keeps your tasks moving forward while you sleep, eat, travel, or just step away from the keyboard. It picks tasks off your queue, isolates each one in a git worktree, watches your Claude Code quota like a hawk, and at the end gives you a per-task report you can review in 30 seconds.

16
0
80% credibility
Found May 23, 2026 at 16 stars -- GitGems finds repos before they trend. Get early access to the next one.
Sign Up Free
AI Analysis
Python
AI Summary

Vigil is a tool that runs Claude Code unattended while you step away. You leave it a list of tasks (like 'fix the login bug' or 'add unit tests'), and it works through them one by one in separate workspaces, automatically pausing when Claude's usage limits are reached and resuming later. Each task produces a clean branch with a detailed report explaining what was done, any decisions made, and what to watch for when merging. You come back to a stack of finished work ready to review.

How It Works

1
📋 You write down your tasks

Before leaving, you tell Vigil what you need done — one line per task. It keeps the list and waits for you to say go.

2
🛌 You close your laptop and leave

You say 'start' and walk away. Vigil is now watching your queue while you sleep, eat, or travel.

3
Claude works through your list — on its own

Each task runs in its own separate workspace. Vigil watches Claude's energy meter and automatically pauses if it needs to rest, then picks back up when it's ready.

4
💾 Every result gets saved automatically

When each task finishes, Claude's changes are committed as a fresh branch — your original project is never touched.

5
You come back to a stack of results
💻
Use the command line

Quick summary table showing which tasks finished, which need review, and what each one cost.

🌐
Open the web dashboard

A colorful page where you can see everything at a glance, click into any task, and manage the queue from your browser.

You decide what to keep

Want the changes? Merge the branch like any other. Not right? Discard it. Either way, the work is ready for you — not lost in a terminal you closed hours ago.

Sign up to see the full architecture

4 more

Sign Up Free

Star Growth

See how this repo grew from 16 to 16 stars Sign Up Free
Repurpose This Repo

Repurpose is a Pro feature

Generate ready-to-use prompts for X threads, LinkedIn posts, blog posts, YouTube scripts, and more -- with full repo context baked in.

Unlock Repurpose
AI-Generated Review

What is claude-vigil?

Vigil is a Python task queue for Claude Code that lets you leave a list of coding tasks before stepping away. While you sleep, eat, or travel, it picks tasks off the queue, runs each one in an isolated git worktree, and monitors Claude's API quota to pause before hitting limits. When tasks complete, you get auto-committed branches and per-task markdown reports summarizing what was done. The CLI handles the full lifecycle: add tasks, start the runner, review results, and merge or discard branches.

Why is it gaining traction?

The killer feature is "vacation mode" - batch autonomous work without babysitting. Vigil handles the friction that usually makes this painful: quota monitoring via a PTY sidecar that captures real 5h/7d usage numbers, automatic retry with exponential backoff for network blips, and a scope hook that blocks Claude from writing outside declared file patterns. The self-summary feature generates structured reports (decisions made, assumptions, merge risks) that take 30 seconds to scan. Dependencies between tasks are supported via a DAG-aware scheduler, and results auto-push to origin.

Who should use this?

Solo developers and small teams running Claude Code who want to batch repetitive work: refactoring, test scaffolding, README updates, or spike investigations. Particularly useful when you have a queue of "fix this, then that" tasks you want processed overnight. Not suitable for environments requiring human-in-the-loop approval or teams uncomfortable with autonomous agent execution writing to repositories.

Verdict

With only 16 stars and a credibility score of approximately 0.8%, this is a nascent project still in dogfooding phase. The concept is solid and the implementation shows thoughtful design around quota handling and crash safety. However, it's pre-release with APIs that may shift before v1.0. Worth evaluating for the "leave tasks, come back to branches" use case if you're comfortable on the bleeding edge, but don't bet production workflows on it yet.

Sign up to read the full AI review Sign Up Free

Similar repos coming soon.