convez.ai
AgenteAutomaçãoIntermediário
coder-zkl1988

Tabby — Desktop AI para Agentes Locais

coder-zkl1988/tabby

Workspace desktop local para agentes IA multi-canal com controle total

Ver no GitHub
BomPluginScore · de 0 a 100

Ambiente desktop local para criar, gerenciar e conectar agentes de IA a canais de chat (WeChat, Feishu, Slack, Discord) mantendo dados e estado de execução no próprio computador. Integra suporte a controle de dispositivos Android, skills customizáveis e automações agendadas.

Instalação

$npx convez-ai coder-zkl1988/tabby

1 comando no terminal: detecta o formato do repositório (marketplace ou plugin avulso) e instala via Claude Code. Requer Node 18+ e o Claude Code instalado.

Métricas

Estrelas
3
Forks
1
Acompanhando
0
Saúde
100%
Último commit
há 7 horas
Versão
v0.9.0
Issues abertas
0
Contribuidores
16

Casos de uso

  • Conectar agentes IA a múltiplos canais de comunicação simultaneamente
  • Automatizar tarefas recorrentes com controle de dispositivos Android
  • Gerenciar modelos IA com BYOK e OAuth sem dependências na nuvem

Veredito

Tabby resolve um problema legítimo: centralizar agentes IA em ambiente local com suporte a múltiplos canais. Arquitetura de desktop-first é adequada para equipes pequenas que valorizam soberania de dados, embora ainda em fase inicial (v0.3.0, apenas ARM macOS).

README

A tradução para o português deste README ainda está na fila.

Tabby

Tabby is a desktop-first AI workspace for running local AI partners, connecting them to chat channels, and controlling device workflows from one application.

Tabby is built for individuals and small teams who want a practical local control plane: create AI partners, connect channels, use your own model providers, and keep runtime state on your own machine.

Download

The current public release is available from GitHub Releases:

Intel macOS and Windows packages are not included in the current release.

What Tabby Does

Tabby provides a local desktop environment for:

  • Creating and managing AI partners
  • Connecting AI partners to chat channels such as WeChat, Feishu, Slack, and Discord
  • Running local OpenClaw-based runtime services from the desktop app
  • Managing model providers, including OAuth-based and bring-your-own-key flows
  • Installing and using skills and expert templates
  • Controlling Android devices and viewing real-time device mirrors
  • Running scheduled and automated tasks

Highlights

Local-First Desktop Runtime

Tabby runs the controller, web UI, and OpenClaw runtime from the desktop app. User configuration and runtime state are stored locally, so your data and automation workflows stay under your control.

AI Partners and Experts

Create custom AI partners for different roles, install expert templates, and use structured workspace files to give each partner a clear identity and task context.

Chat Channel Integration

Connect AI partners to IM channels and make them available from the tools you already use. Tabby includes channel setup flows and bot binding so each channel can be routed to the right AI partner.

Device Control

Tabby includes Android device control and real-time mirroring support. You can connect devices, view live screens, dispatch tasks, and inspect task history from the desktop dashboard.

Skills and Automation

Install skills, sync runtime configuration, and schedule recurring automation tasks. Tabby is designed to move from one-off chat commands toward repeatable agent workflows.

System Requirements

  • macOS 12 or later
  • Apple Silicon Mac for the current arm64 release
  • pnpm 10+ and Node.js 22+ for local development

Installation

  1. Download tabby-0.3.0-arm64.dmg from the release page.
  2. Open the DMG.
  3. Drag Tabby.app into Applications.
  4. Launch Tabby from Applications.

The macOS package is signed with Developer ID, notarized by Apple, and stapled before release.

Development

Install dependencies:

pnpm install

Start the local desktop stack:

pnpm dev start

Stop the local desktop stack:

pnpm dev stop

Run common checks:

pnpm typecheck
pnpm lint
pnpm test

Build the macOS Apple Silicon production package:

pnpm dist:mac:production:arm64

Repository Layout

apps/
  controller/   Local control plane and HTTP API
  desktop/      Electron desktop shell and packaged runtime
  web/          React dashboard
packages/
  shared/       Shared schemas and types
  slimclaw/     OpenClaw runtime packaging contract
tests/          Integration and regression tests
specs/          Product, runtime, and architecture notes

Release Notes

See the latest release notes on the GitHub Releases page.

Acknowledgements

This repository is based on foundational work from the Nexu project; thank you to Nexu for the core groundwork that made Tabby possible.

License

This project is released under the MIT License.

Ver o original no GitHub

Bom para

Relacionados