LeapLab has launched Cooragent, an open-source framework designed for collaborative AI agents. With just one simple prompt:
“Create an AI intelligence-gathering assistant to collect the latest updates in AI for me.”
You’ll have your own dedicated AI secretary that browses the web daily, gathers key insights, and sends you summarized reports with visuals and text. Agents can work together in customizable workflows, unlocking endless possibilities. You can even publish your own agents to the community and share them with others.
🔗 Project Repository: https://github.com/LeapLabTHU/cooragent
Core Philosophy: AGI for Human Collaboration
Traditional agents are often hard to use, requiring complicated setup and customization, making them impractical for everyday use. Cooragent simplifies this with a conversational interface that creates collaborative and editable agents. It is, in essence, an editable AGI:
Collaborative: Agents can work with humans or other agents to complete complex tasks.
Editable: Prompts and behaviors can be modified on the fly.
Prompt-Free: The system understands context dynamically and generates high-precision task instructions automatically—no prompt engineering required.
Two Working Modes
- Agent Factory Description-driven: Just one sentence is enough. The system analyzes your intent and generates a matching agent.
Planner Assisted: The built-in Planner selects tools and optimizes prompts for the agent.
Editable Output: You can fine-tune the agent after generation.
- Agent Workflow Goal-driven: Define the objective, and the system will analyze it and generate subtasks.
Multi-agent Collaboration: The Planner combines agents with domain expertise and orchestrates task steps.
Task Dispatching: Tasks are distributed across agents, which work in parallel to complete them collaboratively.
Prompt-Free Design
Pain Point: Crafting effective prompts is burdensome and time-consuming.
Solution: Cooragent adopts a Prompt-Free design. Through multi-agent collaboration and contextual awareness, it generates optimal instructions without relying on static prompts.
Top comments (0)