For educational and research purposes only — not investment advice.
Trading Agents Lab

Frequently Asked Questions#

Posture, license, what TradingAgentsLab is and is not, and the relationships with upstream and Clawless.


Purpose and posture#

What is TradingAgentsLab for?#

TradingAgentsLab is an educational research tool and paper-trading desktop app. It runs a multi-agent AI debate to produce trade recommendations on any ticker you specify.

For educational research and paper trading. This is not investment advice.

The app is designed to help you understand how AI agents reason about market data, to practice paper trading, and to build intuition for multi-agent analysis. It is not designed to drive real-money trading decisions.

Can I use it to trade real money?#

You can connect a live Alpaca account, but:

  • Live trading is intentionally restricted in the current distribution.
  • The app produces AI-generated recommendations, not financial advice.
  • Any real-money trading decision is entirely your own. TradingAgentsLab is not a registered investment advisor.

Does it guarantee profitable trades?#

No. The agent debate produces a recommendation — BUY, SELL, or HOLD — with a confidence level. The confidence is a model output, not a probability of profit. Past analysis results have no predictive value for future returns. This is a research tool, not a trading signal service.


What TradingAgentsLab is not#

  • Not a brokerage. It does not execute trades on its own.
  • Not a registered investment advisor. Nothing it produces is investment advice.
  • Not a Clawless extension, plugin, add-on, or integration. It is a standalone product that can optionally connect to a Clawless gateway — the same way it can optionally connect to Alpaca or Yahoo Finance.
  • Not a live trading platform. Paper trading is the intended use case.

License#

What license is TradingAgentsLab under?#

TradingAgentsLab additions (the desktop app and engine sidecar) are licensed under AGPL-3.0. The upstream Python core (tradingagents/) is Apache 2.0 and remains Apache 2.0.

The LICENSE file contains AGPL-3.0. LICENSE-APACHE contains the Apache 2.0 text. NOTICE records modification and attribution. CLA.md describes the Contributor License Agreement.

What does AGPL-3.0 mean for users?#

You can use, modify, and distribute TradingAgentsLab. If you distribute a modified version, you must make the source code of your modifications available under AGPL-3.0. Running it for your own use (even on a server) does not require publishing source.

What does Apache 2.0 mean for the upstream code?#

The tradingagents/ directory contains code from Tauric Research's TradingAgents, licensed Apache 2.0. It can be used, modified, and distributed freely. TradingAgentsLab preserves this license in LICENSE-APACHE and records modifications in NOTICE as required by Apache 2.0 §4(b).


Upstream relationship#

Where does TradingAgentsLab come from?#

TradingAgentsLab is forked from Tauric Research's TradingAgents, a multi-agent LLM trading research framework. The upstream project implements a full LangGraph-based pipeline with analyst, researcher, trader, and risk-manager agents.

TradingAgentsLab wraps this core in a desktop UI and a FastAPI sidecar, adds key management, data provider integrations, and the Clawless connector option.

Are changes being contributed back upstream?#

Not at this time. TradingAgentsLab is an independent fork under AGPL-3.0 for its additions.


Clawless relationship#

What is the relationship between TradingAgentsLab and Clawless?#

TradingAgentsLab is a standalone trading companion for Clawless. It is an independent product with its own codebase, license, and UI.

Clawless is one of several optional connectors in TradingAgentsLab — alongside Alpaca, Yahoo Finance, and direct LLM provider keys. Connecting the two is optional. TradingAgentsLab works fully without a running Clawless instance.

No code is shared between the two products. Brand-level coherence (compatible dark palette, compatible font choices) is achieved through independent design decisions, not code reuse.

Does TradingAgentsLab require Clawless?#

No. You can run TradingAgentsLab entirely without Clawless. The Clawless connector (Phase 6) adds an optional gateway routing path for LLM calls.


LLM providers#

Which LLM providers can I use?#

Four provider families are wired end-to-end today: OpenAI, Anthropic, OpenRouter, and Google Gemini. All work with API keys; OpenAI also supports OAuth via your ChatGPT subscription. See configuring-llm-providers.md.

Can I use my paid ChatGPT account instead of an API key?#

Yes — for OpenAI only. The OAuth flow routes debates through chatgpt.com/backend-api/codex/responses (the Codex backend, same as the ChatGPT web app and Codex CLI), so debates bill against your ChatGPT subscription rate limits rather than per-token API charges. See oauth.md. Free-tier ChatGPT accounts are unreliable here — paste an API key instead.

Why is there no Anthropic OAuth?#

Anthropic's Terms of Service prohibit OAuth flows for their API. TradingAgentsLab respects this — Anthropic is API-key only.

Can I use multiple providers simultaneously?#

You can have keys (and OAuth) for all four providers connected at once. Each individual debate uses one provider, picked via the "Run with" dropdown on the Analyze page. Your last selection persists per (provider, auth-mode) combination.


Data and privacy#

Where does market data come from?#

Yahoo Finance by default (free, no API key required). Optionally Alpaca for power users with an Alpaca subscription. See data-providers.md.

Does TradingAgentsLab send data to any server?#

  • Market data is fetched from Yahoo Finance's public endpoints.
  • LLM calls go to your configured provider (OpenAI, Anthropic, OpenRouter, Gemini) using your own key — or, for OpenAI OAuth, through the Codex backend using your ChatGPT subscription.
  • No data is sent to TradingAgentsLab's own servers — the app has none. No telemetry, no analytics, no error reports. All processing happens locally.

Are my API keys safe?#

Yes. Keys are encrypted by your OS keychain (Keychain on macOS, DPAPI on Windows, libsecret on Linux) before touching disk. Plaintext is never written anywhere. OAuth tokens are encrypted the same way. See security-and-storage.md.

Where is my debate history stored?#

Locally, in <repo>/data/sessions.db (SQLite). The History page reads from this file. Delete the file to wipe history. It is not encrypted at rest — store the repo on an encrypted volume (FileVault, BitLocker, LUKS) if you want at-rest encryption for transcripts.


Further reading#