The AI-native workspace for multi-project development.
One dockable window for terminals, AI agents, an editor, Git and pull requests — across local, WSL and SSH projects.
| Download · English | 简体中文 |
Jump to — Why · Preview · Capabilities · Comparison · Use cases · Install · Platform · Develop · Docs · Community · License
Driving several projects with AI agents is a coordination problem, not a typing problem. The pain is never "I can't type the command" — it's "which terminal was that agent in?", "did my commit land?", "the server copy of this repo is 400 files behind".
Neeko is a single workspace that solves those three things:
- Sessions that survive. Every project keeps its own live PTY. Switching projects, switching tabs or restarting the app never kills a running agent.
- One context for code and chat. Terminal, editor, diff view and agent conversation all point at the same working directory — no more hunting for the right tab.
- Local, WSL and SSH behave the same. A folder on a remote server gets the same panels, the same LSP, the same Git workflow as a local directory.
Left: the project sidebar — each project, branch and worktree is a first-class entry. Right: the file tree, decorated with live Git status. Bottom: the status bar — branch, language server, panels, notifications.Not another terminal emulator. Neeko is the layer above your editor and around your agents: it keeps them alive, keeps them visible, and keeps them pointed at the right project.
- Local, WSL and SSH projects — one sidebar for local directories, WSL distributions and SSH hosts, browsed and mounted in place
- Worktree sessions — every git worktree gets its own terminal and working directory
- Restart recovery — project list, selected agent/IDE, worktree state and dock layout are restored
- Sidebar upkeep — drag to reorder, rename branches and worktrees inline, colour-code project avatars
- Per-project settings — override agent, IDE, appearance and tasks for a single project
- Clone from Git — pull a repository in with streaming progress
- One PTY per project — switching projects or tabs never kills a running agent
- 10 built-in CLI agents — Claude Code, OpenCode, Gemini, Codex, Qoder, CodeBuddy, OMP, Pi, Reasonix and Grok, with install detection and one-click launch
- Custom agents — add your own CLI with custom command, arguments and environment; override the built-ins; toggle chat capability per agent
- CLI hooks — 8 external CLIs (opencode, pi, omp, claude, codebuddy, qoder, codex, dsh) report their run state back into Neeko's notifications and status bar
- Split panes & multi-tab — split a terminal into panes and keep several tabs alive across projects
- TUI theme sync — the app theme is written into OpenCode and Pi config for local, WSL and remote projects
- Multi-agent chat — talk to opencode, Pi or any ACP-speaking agent over 4 transports (opencode serve, Pi RPC, ACP, jsonl)
- Streaming with tool cards — reads, diffs, commands, todos, skills and plans render as reviewable cards
- Approvals and follow-ups — approve tool calls, answer questions, confirm plans
- Models and context — pick a model across providers and watch context usage
- Session resume — list past sessions and continue them on the agent side
- Composer intelligence — attachments, slash commands and
@mention completion
A built-in MockAgent lets you try chat without installing any CLI.
- Files and editor — project tree, multi-tab editing, untitled buffers, external-change detection
- Editor groups — split left/right, drag tabs between groups, pin a tab
- Rich previews — Markdown (GFM, mermaid, PlantUML), HTML, SVG, images and JSON
- Navigation — quick open, global search, symbol navigation, reference peek and a command palette
- LSP for 17 languages — Rust, Go, Python, TypeScript, JavaScript, Java, C, C++, C#, Ruby, PHP, Swift, Kotlin, Lua, Elixir, R, SQL
- Diagnostics loop — Problems panel, quick fixes, automatic import edit on completion accept
- Semantic navigation — go to definition, find references and hover, including virtual documents for Java sources
- Custom servers — register your own language server per file extension, with lifecycle policy and one-click install
- Debug — DAP breakpoints, stepping, variables and call stacks for Go, Java and Rust
- Run & test — launch entry points and run individual test cases for Go, Java, Rust and TypeScript
- Unified console — task output, LSP logs and Git operations share one console
- AI translation — translate comments block by block, or whole Markdown/HTML/TXT documents, using an agent you choose
- Change tree — hierarchical file tree with status decoration, refreshed when files change on disk
- Diff — unified or side-by-side, syntax highlighting, word-level diff, change-block navigation
- Commit and sync — commit (message can be generated by an agent), push/pull, credential prompts
- Branches & worktrees — searchable branch switcher, create/rename/delete, worktree add/remove
- Stash & conflicts — stash management and conflict resolution
- History and pull requests — git graph, cherry-pick, tags; PR list, detail, review and merge through the
ghCLI
- One library — skills, prompts and MCP servers in a single searchable surface
- Marketplaces — install skills from skills.sh and MCP servers from the MCP Registry
- Reusable prompts — insert a prompt into the agent composer or the active terminal
- Embedded browser — open a page beside your code; the element picker copies the selected element's HTML to the clipboard
- Session history — scan agent session files, search them, resume or export
- Notifications — in-app notification centre plus desktop notifications (on macOS they jump back to the project)
- Status bar — branch, language server, problems, debug, conflicts, cursor position and notifications in one strip
5 built-in themes — Dark, Light, One Dark Pro, Claude and Classic Dark — plus custom theme files.
Every team already has a terminal and an editor. The difference is whether they stay in sync while an agent works.
| Neeko | Editor + terminal windows | Shell-only CLI agents | |
|---|---|---|---|
| One PTY per project / branch / worktree, kept alive | ✅ | ❌ | |
| Agent run state visible outside the terminal | ✅ status bar + hooks | ❌ | |
| Streaming agent chat with tool cards and approvals | ✅ built in | ❌ needs an extension | |
| Change tree, diff, commit, stash, PR review | ✅ built in | ❌ manual commands | |
| LSP diagnostics and DAP debugging | ✅ built in | ✅ editor-native | ❌ |
| WSL and SSH projects in the same UI | ✅ | ||
| Workspaces and layout survive a restart | ✅ | ✅ |
Neeko does not replace your editor or your agents — it becomes the desk they sit on.
- Parallel agents on one repo — one worktree per agent, each with its own terminal, chat surface and Git state.
- Debug and test without leaving the editor — DAP breakpoints, stepping, variables and call stacks for Go, Java and Rust, plus inline run and debug actions for Go, Java, Rust and TypeScript.
- Remote development without the friction — SSH into a build box and get the same panels, LSP and Git workflow you have locally.
- WSL as a first-class target — Linux toolchain, Windows desktop.
- A personal AI control room — one place to see which agent is running, what it changed, and what still needs a commit.
Grab the installer for your platform from Releases:
| Platform | Artifact |
|---|---|
| Windows | .exe (NSIS) or .msi |
| macOS | .dmg — drag Neeko into Applications |
| Linux | .AppImage or .deb |
macOS first launch — the build is unsigned. If macOS reports the app as damaged or unverifiable,
runxattr -cr /Applications/Neeko.apponce and open it again.
There is no auto-updater: download a newer release to update.
Prerequisites:
- Node.js 24+
- pnpm 11+
- Rust 1.99+ — the exact toolchain is pinned in
rust-toolchain.toml;rustupinstalls it automatically
Linux system packages:
# Ubuntu / Debian
sudo apt install -y build-essential libwebkit2gtk-4.1-dev libgtk-3-dev \
libayatana-appindicator3-dev librsvg2-dev patchelf
# Fedora
sudo dnf install -y gcc gcc-c++ webkit2gtk4.1-devel libappindicator-gtk3-devel librsvg2-devel
# Arch
sudo pacman -S base-devel webkit2gtk-4.1 libappindicator-gtk3 librsvgpnpm install
pnpm tauri dev # development
pnpm tauri build # packaged build| Capability | Windows | macOS | Linux |
|---|---|---|---|
| Local projects, terminals, editor, LSP, debug | ✅ | ✅ | ✅ |
| WSL projects | ✅ | — | — |
| SSH remote projects | ✅ | ✅ | ✅ |
| Desktop notifications (click-through) | ✅ | ✅ (native) | ✅ |
| Embedded browser panel | ✅ | ✅ | ✅ |
Language servers and debug adapters run from your own toolchain; Neeko detects them and can install the common ones.
pnpm check:quick # default gate: static checks + tests for the files you touched
pnpm check # full local run (also executed by pre-push and CI)- TDD is the working mode: failing test first, then the minimal implementation.
- Contribution flow, commit conventions and review rules: CONTRIBUTING.md.
- Repository conventions for humans and AI agents: AGENTS.md.
| Document | What it covers |
|---|---|
| docs/features.md | The canonical capability inventory — the single source of truth this README summarises |
| docs/ARCHITECTURE.md | Module layout, DTO contracts and data flows |
| CONTEXT.md | Domain glossary (project, worktree, agent, skill, task) |
| docs/keyboard-shortcuts.md | The full, remappable shortcut list |
| CHANGELOG.md | What changed in each release |
| CONTRIBUTING.md | How to contribute and what CI enforces |
- ⭐ If Neeko saves you time, star the repository — it is the single most useful thing you can do for the project.
- Issues — bug reports and feature requests. Use the templates; they carry the details we need to reproduce.
- Pull requests — see CONTRIBUTING.md for the quality gates your change has to pass.
- Changelog — what changed in each release.
- Code of Conduct
Community group — join the WeChat group to learn and discuss
Apache 2.0 — see LICENSE.


