marcus/nightshift

Nightshift uses your leftover Claude / Codex budget to surprise you with useful PRs. Love them or leave them.

What it solves

Nightshift is an automated codebase maintenance tool that finds and fixes common technical debt—such as dead code, documentation drift, and security issues—without requiring manual intervention. It acts as a "Roomba for your codebase," running overnight to identify issues and propose fixes via pull requests, ensuring that maintenance tasks don't get ignored while new features are shipped.

How it works

The tool integrates with AI providers (Claude Code, Codex, and GitHub Copilot) via their local CLIs to analyze repositories. It uses a YAML configuration to manage token budgets, target projects, and task priorities. Nightshift schedules runs (e.g., via cron) and executes specific tasks—like lint-fix or bug-finder—and submits the results as branches or PRs so the developer can simply merge or close them.

Who it’s for

Developers and maintainers of multiple repositories who want to automate the tedious parts of codebase hygiene and technical debt reduction using their existing AI subscriptions.

Highlights

  • Budget-aware: Automatically uses remaining daily token allotments without exceeding a configurable maximum.
  • Zero risk: Never writes directly to the primary branch; all changes are submitted as PRs.
  • Multi-project support: Can be pointed at multiple repositories and automatically knows which tasks to run.
  • Flexible AI Providers: Supports Claude Code, Codex, and GitHub Copilot.

Related

  • Project
    hustcer/deepseek-review

    An AI-powered code review tool that uses DeepSeek models to automate PR reviews via GitHub Actions or local CLI audits.

  • Project
    VictorTaelin/AI-scripts

    A collection of AI-powered command-line tools for coding workflows, including terminal chat, automated code editing, and context-aware refactoring.

  • Project
    HexmosTech/git-lrc

    An AI-powered code review tool that hooks into git commit to catch bugs, security leaks, and technical debt before code enters the repository history.

  • Project
    hyhmrright/brooks-lint

    An AI code review tool that diagnoses architectural decay and technical debt by grounding findings in twelve classic software engineering books.

  • Project
    TokenRollAI/llmdoc

    A persistent engineering context system that helps coding agents understand a repository's architecture and design decisions without rediscovering them every session.