AGENT STUDIO / MAINNET
Define your agent.
Set a goal, choose a model provider, and define the tools your agent may request. Save and edit configurations here. Connect your provider and run a saved configuration with Passport approval. Each run produces text and can read permitted chain context.
Choose an agent job
Five ready-made tasks. Choose a role, enter your model and Passport ID, save, then sign a run. These are templates until you run them.
Test a Passport-gated tool
Read the agent wallet’s mainnet balance. Requires an active passport with READ_CHAIN permission and its agent wallet signature. This performs no trade and spends no ETH.
Connect the agent wallet to test.
Connect and run
Run the saved revision once with your selected model. Requires RUN_AGENT permission and the agent wallet signature. Chain context also requires READ_CHAIN. Output is text; trades and external actions are not executed.
OpenAI: No key connected
Keys are encrypted in server storage and sent only to the selected provider. Never enter a wallet private key.
Save an agent configuration first.
Fine-tune a model
For eligible OpenAI accounts and supported base models. Upload your training JSONL to your own OpenAI account with purpose “fine-tune”, then enter its file ID. Other providers can use an already-trained model ID but cannot submit training here yet.
Uses the saved base model and one training epoch. Requires TRAIN_AGENT permission and the agent wallet. Provider billing and dataset size determine the cost; ETH passport limits do not cap that charge.
Open training dashboard ↗Every action requires review. Choosing a tool here does not grant its on-chain permission. Trade placement and custom APIs still require a supported execution adapter. Runs below do not execute the configuration-only tools.
Explore ecosystem ideas ↗
PASSPORT