Why this exists.
Trading Agents Lab provides a high-quality, professional-grade tool purely for educational purposes. We do not force user adoption and we do not provide trading tools — we provide a free resource for analysis and learning.
It is built and maintained by RBJ Global, an independent software studio building practical AI tools. The codebase is a fork of Tauric Research's TradingAgents with substantial extensions for desktop use, cost discipline, paper-trading-only data adapters, and outbound webhooks.
Positioning — analysis, not execution.
This is the most important sentence on this page:
Trading Agents Lab is an analysis tool, not an execution platform.
The product empowers users with multi-agent LLM analysis to make their own trading decisions. It does not execute trades on a user's behalf — not via paper-trading suites that look like real trading apps, not via a hidden feature flag, not via a future patch.
- No live-trading execution code, ever. Not in the engine, not in the UI, not in distribution builds.
- No full trading suite. No stop-loss management, no sell-order workflow, no order routing.
- Defence in depth via endpoint hard-coding. Where the app connects to Alpaca, it connects only to paper-trading and data endpoints. Live keys for the live endpoint simply have nowhere to go in our code.
- External brokerage via webhooks, on your terms. When a debate finishes, the result can be pushed to your own endpoint, which you wire to your own broker on your own authorised account.
This is a regulatory firewall, not a temporary scope cut. When execution code does not exist in our repository, there is structurally nothing to regulate as a trading platform.
A teaching artifact for Clawdemy.
Trading Agents Lab doubles as a practical case study for Clawdemy.org — a free, open AI-education platform. Students can read the source to learn how to design real multi-agent LLM systems that ship.
We optimise for readable code over clever code. Decisions that surprise a reader carry a comment explaining the why, not just the what. The whole system is meant to be understood, forked, and improved on.
If you're a student, researcher, or educator using this as a teaching reference, we'd love to hear from you. File an issue on GitHub and tell us what you're building.
Business model — free, no monetisation.
Free. Open source. AGPL-3.0. No subscription, no paywall, no premium tier, no enterprise upsell.
We don't collect user data because we don't have a business that requires it. There are no analytics SDKs, no telemetry beacons, no error-reporting services, no install pings, no accounts, no email collection. Read the security posture →
If monetisation is ever considered in the future, the change in business model is itself the regulatory inflection point and we will engage securities counsel before shipping anything.
Family.
Trading Agents Lab is part of a small family of independent, single-purpose, privacy-first software:
- Clawless — operating system for AI. Trading Agents Lab is the standalone trading companion for Clawless; it works perfectly on its own and connects optionally.
- WhisprDesk — privacy-first Mac dictation. No cloud, no transcripts leaving the device.
- Clawdemy — free AI education for the worried generalist.
- RBJ Global — the studio behind the family.