Juggler
A visual AI coding harness
Juggler is a dedicated desktop workbench for interacting with code generation agents, providing a structured and visual environment for complex development tasks. Key features include:
* Persistent, branching conversation trees
* Inspectable tool calls and model transactions
* Finder-style column layout for context and threads
* Support for multiple code models (Claude Code, OpenAI Codex, GitHub Copilot, Gemini, Ollama)
* Local and remote session management
This platform transforms the way developers interact with coding helpers by presenting conversations not as linear logs, but as navigable trees. Users can branch conversations into sub-threads for parallel tasks or alternative approaches, integrating only the relevant outcomes back into the main context. The interface is optimized for long sessions, offering focused views for tool arguments, approvals, results, and detailed inspection of every model transaction, including prompts, messages, and token usage.
Juggler is designed for hands-on development, providing a robust environment whether working locally or remotely on a dev box or server. The server manages the session, ensuring all connected clients remain synchronized, allowing for flexible access from a desktop application, browser, or mobile device. It is free to download, open-source, and requires no account, making it an accessible tool for individual developers and engineering teams seeking enhanced control over their code development processes.