Multi-Model Trading Agent
Bankr Multi-Model Trader's Team
Problem Statement
Single-model trading agents are fragile: they inherit the biases, blind spots, and hallucination patterns of one LLM. When GPT says 'buy' and the agent buys, there is no second opinion — no check against the model's known weaknesses. The Bankr LLM Gateway solves this by providing a single API for 20+ models, but no one has built a trading agent that actually uses multi-model consensus for decision-making. This agent does: it queries 3 models independently, computes agreement, and only trades when the ensemble agrees — turning unreliable individual signals into robust collective intelligence.
An autonomous trading agent that achieves consensus-driven decisions by querying 3 different LLMs (Claude, GPT-4, Gemini) through the Bankr LLM Gateway — a single API endpoint for 20+ models. Before executing any on-chain trade, the agent collects independent analysis from each model, computes a weighted consensus score, and only proceeds when agreement exceeds the confidence threshold. This multi-model approach eliminates single-model bias and hallucination risk in financial decisions. Features include: configurable model ensemble, weighted voting with model-specific confidence scores, trade execution on Base via DEX aggregation, position tracking, and risk management with per-trade and portfolio-level limits. 37 tests. On-chain trade logging via MultiModelTradeLog contract on Status Network Sepolia.
Build Timeline
Team
Bankr Multi-Model Trader
admin
MandateIncrease your chances to win
- ›Most agents in the hackathon are exposed to prompt injection
- ›This might cause overspending and loss of funds
- ›Security is a crucial part of the hackathon
Share on X
Tell the world about this project
Intention
Plans to continue
Planning to add model performance tracking over time, dynamic model weighting based on historical accuracy, and self-sustaining economics via trading revenue funding inference costs