CowAgent
Open-source AI assistant and Agent Harness that plans tasks, runs tools and skills, and grows with memory and knowledge. Supports multiple models, channels, and agents; lightweight, extensible, and installable with one line.
Visit zhayujie/CowAgentOverview
CowAgent is an open-source super AI assistant and Agent Harness. It plans and executes tasks, uses tools and skills, maintains long-term memory and a personal knowledge base, and can form multi-agent teams. It supports multiple model providers and chat channels and can run on a laptop or server.
Key Features
- Task planning that decomposes complex tasks and executes them step by step using tools and skills.
- Three-tier long-term memory with Deep Dream distillation and hybrid keyword plus vector retrieval.
- Personal knowledge base that auto-curates structured knowledge into a Markdown wiki and evolving knowledge graph.
- Skills system with one-click installs from Skill Hub, GitHub, and ClawHub, plus conversational custom skill creation.
- Built-in tools for file I/O, terminal, browser, scheduler, memory retrieval, web search, vision, and native MCP integration.
- Pluggable model providers, multi-channel support across Web and IM platforms, and multi-agent teams with distinct roles.
Use Cases
- Running a personal AI assistant 24/7 on a laptop or server.
- Connecting an AI assistant to Web and IM channels such as WeChat, Feishu, DingTalk, Telegram, Slack, Discord, and QQ.
- Creating multi-agent teams with distinct roles for group-chat collaboration and task delegation.
- Extending workflows with installable or custom Skills and MCP tools.
Getting Started
- Linux or macOS installer: bash <(curl -fsSL https://cdn\.link\-ai\.tech/code/cow/run\.sh)
- Windows PowerShell installer: irm https://cdn\.link\-ai\.tech/code/cow/run\.ps1 | iex
- Docker deployment: curl -O https://cdn\.link\-ai\.tech/code/cow/docker\-compose\.yml then docker compose up -d
- Open http://localhost:9899 to access the Web console.
- Download the CowAgent Desktop client for macOS or Windows; the backend is bundled.
Deployment & Requirements
- Linux, macOS, Windows, and Docker deployment modes are supported.
- The one-line installer handles dependencies, configuration, and startup.
- For server deployment, set web_host to 0.0.0.0, set web_password, and open port 9899 in the firewall or security group.
Before You Adopt
- License: MIT. Review its terms before using, modifying, or distributing the project.
- Agent mode can consume substantially more tokens than regular chat, so users should choose models that balance quality and cost.
- The Agent has access to the local operating system, so it should only be deployed in trusted environments.
