Quick Answer
Agent Playground is a free visual canvas for building multi-agent AI workflows. Drag agent nodes onto the board, connect them into reasoning chains, and run the whole graph step by step. It works with free Mistral models out of the box and can also use local Ollama models for fully private inference.
How It Works
Open the playground — it loads with a fresh canvas running on Mistral models
Add an agent node and configure its system prompt and model in the config panel
Connect agents so the output of one becomes the input of the next
Run the workflow and follow each agent’s output as it executes through the chain
Key Facts
- Visual canvas: place agent nodes, wire them together, and watch data flow through the graph
- Runs on free Mistral models by default — connect Ollama to mix in local, private models
- Projects are saved in your browser, so you can close the tab and pick up where you left off
- Configure each agent separately: system prompt, model, temperature, and more
- No sign-up required — the playground opens ready to use
Frequently Asked Questions
Is Agent Playground free to use?
Yes. The playground uses Mistral’s free models by default, so you can build and run multi-agent workflows without paying anything or creating an account.
Can I use my own local AI models?
Yes. If you run Ollama on your computer and it is connected, its models appear alongside the Mistral models and can be selected per agent node.
Where are my playground projects saved?
Projects are stored locally in your browser. They never leave your device and are not uploaded to any server.