The friction between spotting a UI glitch and filing a proper bug report just got significantly lower. ScreenClip 1.0.0, a new macOS menu bar application, launched today with a singular focus: let developers capture a screenshot, annotate it, and hand off the fix entirely to an AI coding agent. The tool, developed by indie hacker Ludo Landry, integrates directly with Codex, Claude Code, and Grok Bot to generate pull requests automatically, bypassing the traditional ticketing workflow that often slows down iteration.

The Workflow: From Capture to PR

The core mechanism relies on a global hotkey, βŒ˜β‡§2, which allows users to select a region of any Mac app or website. Once captured, the user adds a brief note or markup, and ScreenClip routes this context to the selected agent. For local agents like Codex and Claude Code, the tool operates within an isolated Git worktree on the user’s machine. It does not merely suggest a fix; it runs the project’s validation command and opens a GitHub pull request. If automatic merge is enabled, the PR merges only after all checks and review rules pass, notifying the user when it is time to retest.

Agent Compatibility and Remote Options

ScreenClip distinguishes between local and remote agent execution to accommodate different security and infrastructure preferences. Codex and Claude Code run locally, ensuring code remains on the developer's machine until the PR is opened. In contrast, Grok Bot operates remotely via a webhook. When using Grok, ScreenClip pushes approved screenshots to a branch in the project’s GitHub repository, and the remote agent opens the PR. This flexibility allows teams to choose between local processing for sensitive codebases and remote execution for distributed workflows, though local agents require the Mac to remain awake during processing.

Privacy, Licensing, and Requirements

The tool is source-available under the PolyForm Noncommercial 1.0.0 license, making it free for personal use while requiring commercial licensing for enterprise adoption. Privacy is handled through a strict redaction model: users can draw opaque black rectangles over sensitive data, and the unredacted original crops are discarded before sending. ScreenClip itself has no server; data flows directly to the chosen agent’s infrastructure and GitHub. Requirements are specific: macOS 14 or later, a local Git repository, and the GitHub CLI for PR creation. The app is built with Xcode 16 and is available on GitHub for those who prefer to compile from source.

Key Takeaways

  • ScreenClip 1.0.0 automates the full bug-fix loop from screenshot to merged PR for macOS developers.
  • It supports local execution via Codex and Claude Code, and remote execution via Grok Bot webhooks.
  • The tool is free for noncommercial use under a source-available license, with commercial options available.
  • Privacy is maintained via local redaction tools that discard unredacted data before transmission.

The Bottom Line

This is a pragmatic tool for the 'vibe coding' era, removing the bureaucratic overhead of bug tracking. If you’re already using AI agents for development, ScreenClip is the missing link that keeps you in flow state.

Technical Implementation Details

Under the hood, ScreenClip requires careful setup to function correctly. Developers must sign in to GitHub using the gh auth login command to grant permissions for opening pull requests. In the Settings menu, each project must be registered with its local repository path, base branch, and validation command. This ensures that when an agent submits a fix, it is tested against the project’s specific build requirements before a PR is opened. The integration with GitHub is deep; screenshots and agent-generated before/after images are committed directly into the PR, providing visual proof of the fix for reviewers.

Community and Support

As a single-developer project, ScreenClip relies heavily on community feedback for improvement. Landry explicitly encourages bug reports and pull requests on GitHub, noting that the tool is 'free for noncommercial use' and funded by optional tips. The source code is available for inspection, aligning with the transparency expectations of the open-source community. While the Hacker News launch post currently has low engagement, the utility of automating the 'see it, fix it' loop for AI-assisted development makes this a tool worth watching for teams already invested in the Codex or Claude ecosystems.