Introduction
A technical guide to building, evaluating and sharing multimodel trading agents.
What is AgentRouter?
AgentRouter turns a trading strategy into a structured agent blueprint. An agent assigns AI models to research, reasoning, decision and execution-planning roles, alongside explicit market, integration and risk settings.
The builder helps you configure this blueprint through a guided conversation. Saving a blueprint creates a private draft in My agents. Testing, publishing and renting are separate actions with server-side eligibility checks.
System boundaries
| Layer | Responsibility |
|---|---|
| Builder | Collect market, strategy and risk preferences; recommend models and explain each role. |
| Inference | Run chosen models through OpenRouter, including optional prompt-based Auto Router selection. |
| Market data | Retrieve supported public feeds. A data feed does not authorize trading. |
| Execution | Validate supported trade requests and prepare swaps for wallet approval. |
| Marketplace | Store drafts, enforce listing eligibility and verify time-limited rental access. |
Where to begin
Start with Quickstart to build your first blueprint. Architecture explains the request lifecycle. Model routing covers fixed model assignments and OpenRouter Auto Router. Integrations and Rentals document the available capabilities and payment flow.