infiAgent
ActiveDescription
Build your own Cowork, AI Scientist and other SoTA Agents just by editing config files. Support anthropic skills. An infinite-horizon agent framework designed for long-running, complex tasks.
Key Features
- Supports unlimited-duration complex tasks running for days without context accumulation or compression degradation, with full recovery via Resume on any interruption
- Compatible with the Agent Skills open standard — drop skill folders into the library for automatic discovery, loading, and execution by agents
- Flexible agent architecture: multi-level hierarchy (tree-structured orchestration for complex domain tasks) and flat architecture (single agent for general tasks)
- File-directory-based persistent memory system — agents remember all historical tasks across sessions in the same workspace without external databases
- Switchable Thinking/ReAct execution modes with historical task records indexed in local SQLite database and built-in task_history_search tool
- Docker Web UI with multi-user registration and account management, Mac desktop app running fully locally with official skill marketplace
Use Cases
💡 Use Researcher config for long-running academic research with automatic paper generation over multiple days
💡 Build general AI assistants with OpenCowork config for daily tasks and multi-turn conversations
💡 Integrate third-party skills via Agent Skills standard to extend agent capabilities in specialized domains
💡 Deploy multi-user Web UI in Docker environment for team-shared agent capabilities
💡 Build multi-bot collaboration workflows with CheapClaw SDK integrated with IM platforms
Categories
Quick Start
Install: git clone https://github.com/polyuiislab/infiAgent then pip install -r requirements.txt. Docker quick start: docker pull chenglinhku/mlav3:latest, mount ~/.mla_v3 and expose port 4242. Mac desktop available from GitHub Releases. Configure llm_config.yaml with model API, choose Researcher or OpenCowork system to begin.