AI Agent
Trading
Build and deploy autonomous trading agents on prediction markets. Pots Market provides the execution infrastructure — you bring the strategy. Agents run 24/7 in isolated sub accounts with no device required to stay online.
- Self-service: you build, configure, and deploy — Pots Market provides the infrastructure
- Bring your own LLM API key — or rent compute from the Pots Market resource pool
- Install modular skills: data feeds, analysis, execution, and integrations
- Backtest strategy before deploying — agents run 24/7 in an isolated sub account
Four Agent Components
You provide the LLM key, the skills, and the strategy logic. Pots Market provides the execution infrastructure and skill marketplace.
Connect OpenAI, Anthropic, Google, or any compatible provider. You control the model, prompts, and cost. Or rent compute time from users sharing idle API capacity via the Pots Market resource pool.
Modular capabilities that extend your agent. Data skills: real-time orderbook, price feeds, on-chain events. Analysis: technical indicators, sentiment scoring. Execution: place orders, manage positions, stop-loss. Integrations: external data sources, webhooks.
Your agent maintains private local memory: market context, historical observations, strategy state, position tracking. Memory stays private to your agent instance — not accessible to Pots Market.
Write trading logic in plain natural language ("buy Yes when price drops below 30¢ and volume is above $5k") or structured code. Backtest against historical data, iterate on performance, then deploy.
Infrastructure Properties
Agents run on Pots Market's servers — co-located near Polymarket's matching engine in eu-west-2. This matters for short-duration markets.
Compute Resource Sharing
Users with idle LLM API keys can make them available to other users through the Pots Market resource pool. List unused API capacity, other users rent compute time, you earn credits.
This is a compute resource marketplace — similar to how cloud providers offer shared infrastructure. Users accessing shared resources pay for compute usage, not for trading advice or signals.