Welcome. This guide assumes zero prior Python or trading-engine experience. We are going to:
- Get your computer ready (one-time, ~10 minutes)
- Run the existing test suite to prove everything works
- Decide together which broker/data accounts you'll sign up for, and when
You will NOT be writing any code. Claude writes the code. You read it, ask questions, and approve.
Open Windows File Explorer and navigate to:
AlgoTrading - Claude\AlgoTrading\alpha_factory\scripts\
Right-click setup.ps1 → "Run with PowerShell".
A blue terminal window will open and do four things:
- Install uv (a tiny Python package manager from the makers of Ruff)
- Download Python 3.12 if you don't have it
- Install all AlphaFactory dependencies (~150 MB, 2-5 min)
- Run the test suite
You may see: "running scripts is disabled on this system". To fix it ONCE:
- Search the Start Menu for "PowerShell"
- Right-click → "Run as Administrator"
- Paste:
Set-ExecutionPolicy -ExecutionPolicy RemoteSigned -Scope CurrentUser - Press Enter, type Y, press Enter
- Close that PowerShell window and try Step 1 again
At the end you'll see a green box that says "Setup complete. Tests passed."
If you see a yellow box ("some tests failed") instead, scroll up in the terminal, copy the red error text, and paste it into our chat. Claude will fix it.
Once setup is done, just tell Claude:
"Setup script done, tests green."
(or paste any errors). That's it. You're at the end of Phase 1 — you have a tested, working indicators + regime classifier system on your machine.
Double-click scripts\run_tests.bat any time you want to re-run the test suite. No PowerShell, no commands.
You do NOT need any accounts to do Phase 1 (what we just built). The test suite runs entirely offline on synthetic data.
For Phase 2 (real data ingestion — next sprint), you'll need:
| Service | Purpose | Cost | When |
|---|---|---|---|
| Polygon | US equities historical + intraday | ~$30/mo Starter; or $200/mo Stocks Developer for tick data | Sprint 2 |
| Alpaca | Paper trading broker (later live) | Paper = free. Live = $0 commissions | Sprint 4 |
When the time comes, Claude will:
- Tell you exactly which signup page to open
- Tell you which subscription tier to pick
- Tell you what to paste into your
.envfile
You'll never have to figure out what an API key is or what "REST endpoint" means.
Inside this chat, Claude can:
- Write and edit every file in
alpha_factory/ - Read your results, errors, and logs
- Design strategies, run analyses, generate reports
- Tell you exactly what to click next
You step in only when:
- A file needs to run on YOUR computer (we just did this — setup.ps1)
- An external service needs YOUR credentials (API signups)
- A real-money decision is being made (going from paper to live)
- You want to override a recommendation
Everything else is hands-off. Just paste me terminal output when I ask for it, and approve when I ask "ready to ship Sprint 2?"
- Don't run live trades from this code. We are months away from that. The risk policy and code both block it.
- Don't manually edit code unless Claude tells you what and why. Diffs without context create silent bugs.
- Don't paste API keys into the chat. Paste them into your
.envfile (Claude will tell you when). - Don't open Claude Code CLI or use
/goalmode for this project. You're in the right place already — Cowork desktop chat.
When you're ready, run setup.ps1 and tell me what you see.