Repolode
Submit
zhayujie

zhayujie/CowAgent

Open-source personal AI assistant and Agent Harness that plans tasks, runs tools and skills, and keeps memory across web and IM channels.

Open-source alternative to:ChatGPTManus

View on GitHubWebsite
Stars
47,472
Forks
10,398
Open issues
22
Last commit
12 hours ago

What is it?

CowAgent is a self-hosted personal AI assistant that breaks complex requests into steps, loops over tools until the goal is reached, and can control the local computer and external services. It keeps a three-tier memory (conversation context, daily notes, long-term memory) and a topic-organized Markdown knowledge base, with a Web console for chatting, configuring models, and managing skills.

The assistant works with a wide range of LLM providers, including Claude, GPT, Gemini, DeepSeek, Qwen, GLM, Kimi, and local or proxy models, and each of chat, vision, image generation, speech, and embeddings can be routed to a different vendor. It can run 24/7 on a personal computer or server and reaches users through the Web, WeChat, Feishu, DingTalk, WeCom, QQ, Telegram, Slack, Discord, and other channels.

Extensibility comes from installable skills, built-in tools such as file I/O, terminal, browser, scheduler and web search, and native MCP server integration. Teams of agents with their own roles and models can collaborate in a shared conversation.

Key features

  • 01Plans multi-step tasks and loops over tools until the goal is reached
  • 02Three-tier memory with nightly distillation and hybrid keyword and vector retrieval
  • 03Auto-curated Markdown knowledge base with a knowledge-graph view
  • 04Skills installable from a marketplace, GitHub, or created through conversation
  • 05Native MCP integration plus built-in browser, terminal, and scheduler tools
  • 06Connects to Web, WeChat, Feishu, DingTalk, Telegram, Slack and more channels

Use cases

  • Running a personal assistant on a home computer or server that answers through WeChat or Feishu
  • Giving a team of agents shared conversations to split and complete larger tasks
  • Automating scheduled jobs and file or web workflows from chat
  • Trying out multiple LLM providers behind one assistant interface

Quick start

  1. 1

    Run the one-line installer on Linux or macOS

  2. 2

    Or run the Docker setup (download compose file, then start)

    curl -O https://cdn.link-ai.tech/code/cow/docker-compose.yml
  3. 3

    Start the containers in the background

    docker compose up -d
  4. 4

    Open the Web console at localhost port 9899

Things to know

  • Install relies on a remote script, so review it before running on a machine you care about.
  • Most features need an API key from a third-party LLM provider.

Topics

#ai-agent#llm#personal-assistant#mcp#multi-agent#chatbot#skills#self-hosted#ai#ai-agents

FAQ

What is CowAgent?
Open-source personal AI assistant and Agent Harness that plans tasks, runs tools and skills, and keeps memory across web and IM channels. CowAgent is a self-hosted personal AI assistant that breaks complex requests into steps, loops over tools until the goal is reached, and can control the local computer and external services. It keeps a three-tier memory (conversation context, daily notes, long-term memory) and a topic-organized Markdown knowledge base, with a Web console for chatting, configuring models, and managing skills.
Is CowAgent free to use? What license does it have?
CowAgent is open source under the MIT license. Check the license terms before using it in commercial products.
Is CowAgent still maintained?
Actively maintained. The last commit to CowAgent was on Oct 20, 2026.
What can I use CowAgent for?
Running a personal assistant on a home computer or server that answers through WeChat or Feishu Giving a team of agents shared conversations to split and complete larger tasks Automating scheduled jobs and file or web workflows from chat Trying out multiple LLM providers behind one assistant interface
How do I install CowAgent?
Run the one-line installer on Linux or macOS Or run the Docker setup (download compose file, then start): curl -O https://cdn.link-ai.tech/code/cow/docker-compose.yml Start the containers in the background: docker compose up -d Open the Web console at localhost port 9899
Where can I get CowAgent?
The source code and installation instructions are on GitHub: https://github.com/zhayujie/CowAgent

Summary generated by AI from the project's README. GitHub stats refresh daily.

Related projects