Nimbalyst is a free, open-source desktop app that gives you a visual workspace for AI coding agents like Claude Code and OpenAI Codex. Instead of running agents in an opaque terminal, you get rich visual editors for markdown, mockups, diagrams, and code, plus inline diff review so nothing changes without your approval. Run parallel agent sessions, organize work on a kanban board, and track tasks in one place. Available on macOS, Windows, Linux, and iOS.

• Inline AI Diff Review: Every change from Claude Code or Codex appears as an inspectable red/green diff. Accept, reject, or edit each change before it touches your codebase.
• 7+ Visual Editors: Edit markdown, CSVs, mockups, Excalidraw diagrams, data models, Mermaid diagrams, and code with full AI context awareness.
• Parallel Session Management: Run 10+ AI sessions simultaneously, organized on a kanban board. Search, filter, tag, and group sessions into workstreams.
• Git Integration & Worktrees: Built-in visual diffs, branch management, and AI-assisted commit messages. Each agent session can operate on its own branch via worktree support.
• Built-in Task Tracker: Track bugs, features, ideas, and projects with status, tags, and priority. Agents update task status automatically as they work.
• MCP Server Integration: Connect agents to Linear, GitHub, Slack, databases, and other services through per-provider MCP configuration.
• Extension Marketplace: Register custom editors for new file types and install community extensions that plug into the agent-aware workspace.
• Mobile App (iOS): Start new sessions, monitor running sessions, review agent work, and chat with Claude Code and Codex from your iPhone.

• Free for Individual Users: The full workspace costs $0/month with no feature limits or trial period, so your only cost is your AI API usage.
• Local-First and Private: All files stay on your local filesystem in standard formats like markdown, code, and diagrams. The desktop app is open source, and data is stored locally with optional usage telemetry controls.
• Visual Diff Review Prevents Mistakes: Every AI-generated change is surfaced for approval before it lands in your repo, giving you full control over what gets committed.
• Multi-Role Workspace: Useful beyond developers. Product managers draft PRDs and mockups, designers build prototypes, and data practitioners clean and document datasets.
• Open Source (MIT License): The desktop app is MIT-licensed, so you can inspect, modify, and contribute to the codebase.

• Bring Your Own AI Costs: Nimbalyst is free, but you pay separately for Claude Code, Codex, or other model access, which can add up at scale.
• Limited Agent Support: The best experience is with Claude Code and Codex. Support for OpenCode, Copilot, and Codex via ACP is still in alpha.
• Learning Curve for Newcomers: Concepts like git worktrees, session kanban boards, and multiple visual editors can feel heavy for users new to agentic workflows.
• Team Features Not Yet Available: The Team plan with multiplayer collaboration is still in development with pricing TBD, so it's currently a solo tool.
• Interface Can Feel Dense: Having mockups, diagrams, code, and sessions in one workspace may feel overwhelming until you settle into a layout that works for you.

• Feature Development with Guarded Diffs: Spin up a Claude Code session, describe the feature you need, and review every code change as a visual diff before it lands in your repo. Run multiple feature branches in parallel using git worktrees without conflicts.
• PRD and Spec Writing for Product Managers: Describe what you need and let Claude Code draft a complete product requirements document with acceptance criteria, edge cases, and relevant file paths from your codebase. Push tickets directly to Linear via MCP integration.
• Architecture and Flow Diagrams: Describe your system in plain language and get instant Excalidraw or Mermaid diagrams. Refine them on the canvas, then use them as context for implementation sessions with your AI agent.
• Bug Investigation and Fixing: Paste a stack trace or error log and let Claude Code analyze the root cause, write a minimal reproduction case, and propose a fix. Review the patch as a diff before merging.
• Mockup and Prototype Creation: Describe a screen or user flow and get a wireframe generated in seconds. Edit directly or ask the AI to adjust spacing, colors, and components. Turn mockups into working HTML/CSS prototypes without writing code.
• Data Cleanup and Documentation: Clean and reshape CSVs, enrich datasets, and document data pipelines inside one AI-aware project space using the built-in visual CSV editor.