Meet the Agents 🤖
Agents are smart AI loops that use tools, remember context, and execute tasks—all optimized for local small models.
Think of a standard LLM as a very smart, but slightly forgetful, static text predictor. You ask a question, it spits out an answer, and immediately forgets you exist.
A Kern Agent is like giving that LLM a brain, hands, and a memory! 🧠 It wraps a stateful control loop around a stateless model. Guide it with instructions, hand it some tools (like web search or database access), and watch it reason, call tools, and get things done in a neat loop.
Because Kern is designed for small language models (1-7B parameters), our agents don't need giant, expensive cloud clusters. They run perfectly on your laptop using Ollama or llama.cpp. 💻
🗺️ What can we build?
Build an Agent 🛠️
Learn how to spawn agents, wire up tools, and set rules.
Run an Agent 🚀
Execute agents, handle responses, and stream tokens.
Debug an Agent 🐛
Troubleshoot model reasoning and inspect execution paths.
👥 Beyond Single Agents
Sometimes, a single agent gets overwhelmed. Don't worry—Kern has abstractions to help them collaborate:
| Abstraction | What it does |
|---|---|
| Team 🤝 | Group of specialized agents working together under a coordinator. |
| Workflow ⛓️ | A predictable, structured pipeline of steps (agents, teams, or functions). |
📚 Resources & Next Steps
- Want to see full code examples? Check out the Agent Examples!
- Ready to look under the hood? Read the Agent API Reference.