Juggler

Juggler reimagines how developers work with AI agents by replacing text-based transcripts with a visual workbench. Built by Julian Storer, it provides a desktop application (macOS, Windows, Linux) and optional remote browser access for collaborative sessions.

Highlights

• Tree-structured conversations instead of linear logs, with full undo/redo and branching navigation
• Complete transparency: view tool calls, approvals, message context, and raw JSON payloads
• Extensible plugin architecture using JavaScript for custom context items, strategies, and commands
• Multi-client support keeps sessions synchronized across desktop and browser simultaneously
• Works with major LLM providers: Claude, OpenAI, Gemini, Ollama, OpenRouter, Deepseek
• Editable context and inline text editing for large code blocks

Juggler appeals to developers who collaborate frequently with AI assistants and value visibility over their agent's reasoning and actions. The visual approach makes it easier to explore alternative branches, review decisions, and maintain control. Completely free and open-source under AGPL-3.0.