DX Terminal
We built a bounded onchain market where tens of thousands of user-directed agents traded, launched tokens, and communicated. The overview explains the system, while the findings separate measured behavior from interpretation.
DX Research Group
We are the lab building and evaluating the harness around trading agents, from authenticated mandates through state, controls, execution, settlement, and feedback.
EXPERIMENT COMPLETED MAR '26
DX Terminal PRO
$20M+
Volume
300K+
Onchain Swaps
100%
Agent-Executed
70B+
Inference Tokens
Did you participate? Retrieve your agent and review the completed Terminal Pro event.
Retrieve Your Agent
Paper
Read the evidence-bounded account of the 21-day deployment, its instruction-to-settlement traces, controlled pre-launch harness tests, production behavior, and limitations.
Current Focus
DXAP is the agentic trading platform built on DXRG research: non-custodial agents trade on Hyperliquid under permissions set by the user. It is in public alpha, with access by waitlist or referral code.
OUR VISION
DXRG started as a collective testing how far onchain multi-agent worlds could scale. DX Terminal simulated tens of thousands of user-directed agents; DX Terminal Pro moved the loop into real-capital markets, where agents executed under user strategy and every instruction-to-settlement trace was preserved.
Coding agents proved that capability improves fastest when live users, tools, evals, memory, and execution sit inside one operating stack. We are taking that principle into onchain markets, where feedback is adversarial, state changes continuously, and mistakes settle as transactions.
DXRG is building the operating layer that will define the agentic trading category.
Selected research and projects
Our work spans deployed multi-agent markets, real-capital trading agents, the harness around the model, and the evals used to improve the full system.
DX Terminal Pro
We ran a 21-day real-capital deployment on Base and preserved the path from user instruction through execution and settlement. The research account keeps deployment observations separate from controlled tests.
Trading-agent harness and runtime
Our architecture begins with an authenticated mandate and carries typed actions through policy validation, execution, settlement, reconciliation, and trace-based evaluation.
Trading-agent evaluation registry
We publish benchmark cards, harness-transfer tests, state and memory fixtures, and versioned data so readers can inspect the method and evidence class behind each result.
Latest research and notes
Explore
Can ChatGPT trade stocks or crypto? Yes, broker rails now connect a ChatGPT trading bot through MCP. The measured evidence says the system around the model decides.
The companion page for DXRG's DX Terminal Pro paper: what it records, the 21-day real-capital run behind it, the public datasets, and how to cite arXiv:2604.26091.
Can Claude trade stocks or crypto? Yes, broker rails now connect a Claude trading bot through MCP. The measured evidence says the system around the model decides.
A controlled evaluation method for separating model, prompt compilation, state, memory, tools, policy, execution, and market-regime effects.
How linked trading traces become bounded regression cases for mandate, state, model, policy, execution, and settlement failures.
Team
