zhayujie/CowAgent

Open-source super AI assistant & Agent Harness. Plans tasks, runs tools and skills, self-evolves with memory and knowledge. Multi-model, multi-channel. Lightweight, extensible, one-line install. (formerly chatgpt-on-wechat)

What it solves

CowAgent is an open-source AI assistant designed to move beyond simple chat by proactively planning tasks, controlling computers and external services, and building a personal knowledge base that evolves over time.

How it works

It operates as an "Agent Harness" where messages enter through various channels (like Telegram or Slack) and are processed by an Agent Core. This core uses a planning system to decompose complex tasks and executes them using a combination of tools (atomic capabilities like file I/O and web search) and Skills (higher-level workflows). It supports a wide range of LLM providers and integrates with the Model Context Protocol (MCP) for extended tool access.

Who it’s for

Users who want a personal, 24/7 AI assistant that can be deployed locally or on a server, integrated into their favorite messaging apps, and capable of performing actual system operations and knowledge management.

Highlights

  • Self-Evolution: Automatically reviews conversations to improve skills and consolidate memory.
  • Three-Tier Memory: Uses a context, daily, and core memory architecture with a "Deep Dream" distillation process to refine long-term memory.
  • Extensible Skills: One-click installation of workflows from a Skill Hub or creation of custom skills via natural language.
  • Broad Integration: Native support for numerous IM platforms (WeChat, Slack, Telegram, etc.) and a wide array of LLM providers.
  • MCP Integration: Supports the Model Context Protocol for zero-code integration of external tools.