An autonomous multi-agent coding framework built on top of Claude Code

You know that feeling when you're juggling multiple coding tasks — fixing a bug while implementing a feature while reviewing a PR — and Claude Code keeps losing context between sessions? You start a task, the agent runs for hours, you come back to find it stuck or the main branch polluted with half-finished code. Managing parallel AI coding sessions manually is tedious: different terminals, different branches, no visibility into what each agent is doing.

You open a git repository in Aperant's desktop app and create tasks on a Kanban board. Each task spawns in an isolated git worktree — a separate copy of your code that shares the git history but has its own working directory. Claude Code agents run in these worktrees, so one agent's changes don't affect another's. The app shows you real-time progress: which agents are coding, which are stuck, which need review. When a task completes, Aperant runs its QA loop to validate the changes, then creates a PR for you to merge.

• Parallel agent execution — Run up to 12 Claude Code agents simultaneously, each in isolated git worktrees, so your main branch stays clean while agents work independently
• Kanban task management — Visual board showing planning, working, in-review, and done columns with real-time progress indicators and file-level status
• Self-validating QA loop — Agents run tests and validate their own changes before marking tasks complete, catching issues before you review
• Persistent memory layer — Stores architecture decisions and coding conventions across sessions so agents follow your project's patterns in future tasks
• GitHub/GitLab/Linear integration — Import issues directly, investigate with AI, create merge requests, and sync task status with your existing workflow tools
• Auto-updates with beta channel — Desktop app updates automatically, with opt-in beta releases for early access to new features
• Cross-platform desktop apps — Native installers for Windows, macOS (Intel and Apple Silicon), and Linux (AppImage, deb, Flatpak)

Who It Is For

If you're a developer who uses Claude Code daily and finds yourself context-switching between multiple features or bug fixes, this multiplies your throughput. Not useful if you work on single-task-at-a-time projects or don't have a Claude Pro/Max subscription (required). Best fit for solo developers and small teams who want autonomous coding without the overhead of enterprise tools.

Worth Exploring?

Yes — the project is mature (37 releases, 1,100+ commits), actively maintained (latest release Feb 2026), and solves a real pain point for power users of Claude Code. The open-source AGPL license means you can inspect every line of code. Start with the stable release (v2.7.6) rather than beta if you're using it for real work.