ИИ-агенты
ECC
Agent harness optimization system with skills, memory and security for Claude Code, Codex and more
★ 275,2 тыс. звёздMIT
Смотреть руководство →open-cowork — проект с открытым исходным кодом из категории «ИИ-агенты» от OpenCoworkAI (OpenCoworkAI/open-cowork на GitHub). У него 2 195 звёзд на GitHub и 310 форков, он распространяется по лицензии MIT и последний раз обновлялся на GitHub 6 октября 2026.
Open Cowork: a desktop AI agent app with one-click Claude Code, MCP tools, skills, and VM sandboxing.
Decide which workspace folder the agent may work in, and which model provider you will connect.
Install the app via Homebrew on macOS, the downloadable installer, or build it from source.
Open Settings, paste your API key, and set the provider base URL and model name exactly as the provider table specifies.
Select a workspace folder and give the agent a task, such as summarising a spreadsheet into a presentation, and watch its work in the trace panel.
Pre-built Windows and macOS installers need no environment setup, and the chat interface supports drag-and-drop files and images.
Works with Claude, OpenAI-compatible APIs, and providers such as GLM, MiniMax, and Kimi, including through OpenRouter, with flexible base-URL configuration.
Commands run in WSL2 on Windows or a Lima VM on macOS when available, with a path-guard fallback that confines file operations to the chosen workspace.
Bundled skills generate and process PPTX, DOCX, XLSX, and PDF files, and users can create custom skills.
Connects to browsers, Notion, and other desktop apps through MCP, and can operate desktop GUIs directly via computer use.
Feishu (Lark) and Slack integration let you direct the agent remotely, while the trace panel shows reasoning and tool calls in real time.
End users can skip this and use the Homebrew cask or installer instead.
git clone https://github.com/OpenCoworkAI/open-cowork.git
cd open-cowork
npm installThese are the documented source-build commands from the README.
npm run rebuild
npm run devOpen Settings in the app, add your provider API key, base URL, and model, then pick a workspace folder and enter a prompt.
npm run build # optional: build the installer locallyIt is maintained by the OpenCoworkAI organisation on GitHub.
Yes, the app is MIT-licensed and free. You pay only your chosen model provider for API usage.
Yes: the agent is powered by your own API key from a supported provider such as Anthropic, OpenRouter, Zhipu (GLM), MiniMax, or Kimi.
Claude Cowork is Anthropic's closed product; Open Cowork is an open-source implementation that adds GUI operation and remote control alongside MCP and skills support, per its own comparison table.
Install via Homebrew (brew install --cask open-cowork after tapping OpenCoworkAI/tap) or the release installer, add your API key in Settings, choose a workspace, and prompt.
Agent harness optimization system with skills, memory and security for Claude Code, Codex and more
Product Management skills framework built on battle-tested methods for Claude Code, Cowork, Codex, and AI agents.
Code, build and evaluate LLM agents with full MCP, Skills, ACP and A2A support.
Rebuild Claude Code’s core in ~5,000 lines: an 11-chapter coding-agent masterclass.
Run context-engineering-kit locally: hand-crafted claude code skills focused on improvi...
A Claude Code-style coding agent that runs anywhere and works with almost any model provider
Другие проекты из категории «ИИ-агенты» с наибольшим числом звёзд в каталоге LibreAI — реалистичный список альтернатив, если open-cowork вам не подходит.
A personal AI that really does things — on any OS, with your data staying yours.
An agentic skills framework & software development methodology that works.
Agent harness optimization system with skills, memory and security for Claude Code, Codex and more
The agent that grows with you — a Python framework for adaptive AI agents.
DeepSeek's plugin-based agent harness where everything is a plugin.
Соседние проекты из категории «ИИ-агенты» в каталоге, ближайшие по числу звёзд к open-cowork — продолжайте изучать полку по обе стороны от него.
Self-hosted AI vulnerability research: agents find, validate, deduplicate, and rank real security issues.
A benchmark for tool-agent-user interaction in real-world domains.
Unla MCP Gateway: turn existing APIs and MCP servers into MCP endpoints with zero code changes.
Shrimp Task Manager: an MCP server that turns natural language into structured, dependent dev tasks.
Agno's self-learning data agent: grounded text-to-SQL answers over your database that improve with use.
One open platform to simulate, evaluate, guardrail, trace, and optimise LLM and agent apps, self-hostable.
Бесплатно и открыто — забирайте исходный код, ставьте звезду репозиторию и начинайте создавать.