Handler

Handler is a macOS desktop application built for developers who use AI coding agents like Codex and OpenCode. It wraps the agent workflow in a review layer where every proposed edit arrives with a built-in explanation and its own isolated side chat, so you can question or reject individual changes without disrupting the main agent context. The app is currently available as a direct download (20.8 MB DMG) at version 0.4.11.

What It Is

Handler sits between you and your AI coding agent, surfacing each proposed file edit as a reviewable unit with an attached explanation. Instead of receiving a 600-line diff and skimming it before accepting, you see each change broken out, can open a side chat to ask why it was made or request a redo, and only then let it land in your codebase. The tool positions itself as a guard against "AI slop" — code you ship but cannot explain.

Core Workflow

The Handler workflow follows a plan-then-review model:

• Plan before writing: Outline the approach with the agent before it writes a single line.
• Per-edit review: Each change comes with its own explanation and a side chat that stays off the main agent thread.
• Terminal awareness: The agent reads the terminal output directly, expanding JSON and pointing to the exact row that caused an error — no manual log-pasting required.
• Screenshot-to-agent: Capture anything on screen and hand it to the agent via the command bar.
• Model switching: Switch between OpenCode and Codex without losing context.
• Isolated worktrees: Spin up separate worktrees so multiple agents never collide.
• Multi-agent tab UI: Run several agents simultaneously in one clean tabbed interface.

Platform and Deployment

Handler is a native macOS desktop application distributed as a direct DMG download. There is no web interface or browser extension — the experience is local and desktop-first. The current stable release is version 0.4.11.

Roadmap Signal

The pricing page notes that the current early bird price will increase when the team ships a "unified memory layer" — described as one shared memory across every agent you run, so none of them start from scratch. This feature is listed as "coming soon" on the homepage, indicating active development toward a more persistent, cross-agent context system.

Who It Is For

Handler targets developers who already use AI coding agents and want more control over what gets committed. The core problem it addresses is the tendency to accept large, opaque diffs from agents without fully understanding the changes — a risk that grows as agents become more capable and produce larger outputs per session.