# Add to your Claude Code skills
git clone https://github.com/BotCoder254/limboolimboo is an open-source ai agents skill for AI coding assistants such as Claude Code, Codex CLI, and ChatGPT, built by BotCoder254. An Orchestrate multiple coding agent desktop app. It has 51 GitHub stars.
limboo's catalog security scan is still queued. You can run an instant dependency and prompt-injection check now with the "Scan for vulnerabilities" button above.
Clone the repository with "git clone https://github.com/BotCoder254/limboo" and add it to your Claude Code skills directory (see the Installation section above).
limboo is primarily written in TypeScript. It is open-source under BotCoder254 on GitHub, so you can review or fork the full source.
Yes. SkillsLLM lists many other AI Agents skills you can browse and compare side by side. Open the AI Agents category from the badge at the top of this page, or use the Related Skills and comparison links further down to weigh limboo against similar tools.
No comments yet. Be the first to share your thoughts!
Unlocks once the catalog security scan passes (runs nightly).
The deep catalog scan for this skill is still queued. Run an instant dependency check now instead.
The operating system for AI software development.
A local-first desktop workspace that gives a connected coding agent everything it needs to do real engineering: projects, sessions, file watching, repository indexing, git, terminals, memory, permissions, and context. Limboo is not an AI model. It is the environment around one.
Documentation · Architecture · Contributing · Roadmap · Security
Limboo is a desktop application (Electron + React + TypeScript) that acts as the workspace around a coding agent. The agent already understands programming, debugging, planning, and git. Limboo provides the environment where it can perform at its highest level: it manages the repository, watches the filesystem, runs the terminal, owns the database, holds durable project memory, and enforces a strict security boundary, while the agent focuses exclusively on writing software.
Every unit of work is a Session — a bundle of a repository, branch, chat history, agent, terminal history, checkpoints, permissions, context, memory, tasks, and generated files. Instead of opening many windows, everything lives inside one workspace.
Traditional IDEs revolve around files. Limboo revolves around conversations. You do not ask "which file should I edit?" — you say "implement authentication," and the agent figures out the files while Limboo visualizes the process: the streaming reply, the tool calls, the file changes, the git diff, and the running commands.
It is local-first by design. There is no backend and no cloud sync. The only network traffic is the connected coding agent talking to its AI provider. The project, its history, and its memory belong to the developer.
#000000 background.Most agent tools stop at "resume the conversation." That restores what the agent said, but not the world it said it in. Limboo treats the environment as a first-class, continuously-maintained system that cooperates with the agent instead of leaving it to rediscover reality on every turn.
The agent resumes against verified repository reality, not a transcript. The Resume Pipeline is the flagship of this idea. When you reopen a session — after an hour or after three weeks of other people's commits, rebases, and dependency bumps — Limboo revalidates the git worktree against the exact state the session last saw and computes a structured repository delta: commits landed, files changed (with dependency manifests and migrations flagged), symbols added or removed, and which files import what changed. That delta is injected once, before your next prompt, so the agent reconciles its assumptions up front instead of burning turns re-reading the tree. It is fully local, uses only bounded argv-only git, and never blocks you from switching sessions. See the Resume Pipeline.
Isolation by default. Every session can own its own git worktree — a real isolated checkout and branch — so parallel work never collides, and "continue where I left off" means a specific filesystem, not a shared one.
The app owns the knowledge, not the model. Durable project memory, the search index, and now a symbol/dependency graph are platform services the app maintains and injects — provider-independent and offline. Memory even references repository symbols, so guidance whose code was deleted is automatically demoted until it reappears.
A hard security boundary you can trust. The renderer performs nothing: all filesystem, git, shell, and database access lives in the main process behind a single typed IPC bridge with sender validation, deny-by-default permissions, and parameterized SQL throughout. Git never runs through a shell; paths are guarded against traversal; secrets are never stored.
The result: continuation feels genuinely intelligent, because the agent begins each resumed task with an accurate, up-to-date understanding of a living codebase.
A tour of the shell: sessions on the left, the conversation in the center, and the activity rail on the right — every panel below is one click on the rail.
| Files — the indexed workspace tree with per-language icons | Changes — live git status with per-file diff stats |
| Git — stage, commit, history, branches, and checkpoints | Memory — durable project knowledge with tiered proposals |
| Tasks — the agent's live plan and progress | Terminal — workspace-scoped PTY sessions |
| Command palette — every command behind Ctrl/Cmd+K | Global search — files, symbols, docs, memory, commits, sessions |
Settings — general, appearance, agent, git, memory, and advanced JSON editing in one modal:
Renderer (Chromium + React) UI only — it asks, it never performs
| window.limboo.*
v
Preload (contextBridge) the only bridge; contextIsolation on, sandbox on
| ipcRenderer <-> ipcMain
v
Main (Node.js + OS access) workspaces, sessions, git, terminal, fs, agent,
memory, SQLite — every OS-touching capability
Limboo runs three Electron contexts with a hard boundary between them. The renderer holds no business logic; all filesystem, git, shell, database, and agent work lives in the main process and crosses a single typed IPC bridge. See docs/architecture/overview.md.
| Layer | Choice |
|---|---|
| Shell / desktop | Electron 42 (via Electron Forge 7) |
| Bundler | Vite 5 (@electron-forge/plugin-vite) |
| UI framework | React 19 |
| Language | TypeScript |
| Styling | Tailwind CSS v4 (CSS-first, no config) |
| State | Zustand 5 (slice-per-domain stores) |
| Database | better-sqlite3 (WAL, FTS5) |
| Terminal | node-pty (Node-API) + xterm |
| File watching | chokidar |
| Coding agent | @anthropic-ai/claude-agent-sdk |
| Icons | lucide-react |
Prerequisites
better-sqlite3 (build-essential / Xcode Command
Line Tools / MSVC Build Tools depending on platform) — it may compile on
install if no matching prebuilt is published. node-pty (pinned to the
Node-API 1.2.0-beta line) ships an ABI-stable prebuilt and never compiles.ANTHROPIC_API_KEY or the
Claude Code credentials file). See
docs/guides/using-the-agent.md.Run in development
npm install # installs deps and compiles native modules
npm start # Electron + Vite dev server (renderer on :5173)
There is no npm run dev — npm start drives both Electron and Vite. Full setup
notes live in [docs/getting-starte