Build an Alphanume + TradeZero Workflow with an AI Coding Assistant

You can ask a coding assistant to help install packages, explain the code and prepare scheduled tasks. Give it one small job at a time and ask for evidence before moving on. This guide uses ordinary prompts you can adapt to your preferred coding assistant.

The assistant is helping you operate the downloaded Python project. These prompts do not require a broker-connected chat tool, and installing an assistant does not give it Alphanume or TradeZero access automatically.

This example requires an Alphanume API key with access to Pre-Market Drop Risk. TradeZero handles the account and execution; Alphanume supplies the ranked candidates. TradeZero's API can use other data, but this starter is built around Alphanume. See dataset access and plans.

Download the Python starter, Windows wrapper, and offline tests. Extract it before running commands. The ZIP includes setup instructions and the complete source.

Before the first prompt

Extract the starter to a permanent Windows folder and open that folder in your coding tool. Have Python 3.11 or newer installed. Obtain the necessary accounts separately: TradeZero API access, Alphanume dataset access, and fresh market-data access for the exit.

Store key values through Windows' user environment variable editor. Give the assistant the variable names only: TZ_API_KEY_ID, TZ_API_SECRET_KEY, ALPHANUME_API_KEY, and MASSIVE_API_KEY. Use TZ_ACCOUNT_ID if account selection needs it. Do not paste secret values into a prompt or ask the assistant to print them.

Prompt 1: inspect the project and install locally

Read README.md, client.py and starter.py in this folder. Explain the
four commands and the two write flags in plain English. Create .venv,
install requirements.txt, and run the offline tests. Do not call a
brokerage API, change strategy rules, or register scheduled tasks yet.
Report the test result and any missing setup. Do not print credentials.

Expected check: a local environment, passing offline tests and a clear explanation of check, entry, exit and reconcile. Test success validates the mocked cases, not live fills or account permissions.

Prompt 2: verify read-only access

Check whether the required environment variable names are set without
printing their values. Explain how to set missing user variables on
Windows. Run starter.py check for my intended Paper environment.
If the account type differs, stop and tell me. Summarize the route
names and Alphanume signal count. Do not submit or cancel orders,
request locate quotes, or accept fees.

Expected check: the intended environment, route names and a data count. If you deliberately use live keys, change “Paper” to “Live” and require --expect Live, keeping all write flags absent. An empty signal list is not a reason to invent tickers or use yesterday's data.

Prompt 3: explain a dry-run entry

Run an entry preview with a fictional $500 reference budget and no
write flags. Show how Alphanume max_rank=5 and min_price=1 affect the
list, and explain each proposed order field. Check the account routes.
Identify candidates needing locates. Explain why the pre-market sizing
price may differ from the opening fill. Do not change the budget or
enable submission.

Expected check: a preview using real entitled Alphanume rows if available, with quantities and borrow checks. No locate quote or order should appear in the broker account because the command makes only GET requests.

Prompt 4: prepare Windows tasks for review

Read scheduled.ps1 and the Windows scheduling guide. Confirm my local
Windows time zone matches Eastern Time with daylight saving. Prepare
the three dry-run task definitions using my actual extracted folder
and virtual-environment Python. Show the weekday triggers, arguments,
working directory, run-as user, overlap setting and log paths. Keep
secrets out of task arguments. Do not register the tasks until I have
reviewed those exact definitions.

Expected check: entry, exit and reconciliation definitions matching the scheduling guide. Check how early closes, sleep and signing out affect them. Once reviewed, you can explicitly ask your tool to register those definitions.

Prompt 5: rehearse recovery before considering writes

Use the mocked tests to demonstrate a rejected order, a timed-out
submission, a partly filled cover, an unconfirmed cancellation and
a stale quote. Explain which IDs are saved and why rerunning entry
does not repeat the basket. Show the explicit Paper submission command
and the separate Live order and paid-locate options for my review.
Do not run any write-enabled command or delete state.

Expected check: a readable recovery explanation and the exact commands you would have to authorize. Ask the assistant to explain any mismatch with the README before you change configuration.

Keep the reasoning tied to evidence

An assistant can write plausible code using the wrong field names or confuse a pending response with a fill. Have it link the official API reference for each changed request and run the offline suite after edits. The downloadable code is the starting point; its tests, saved IDs and broker readbacks are the checks.

Common questions

Can I do this without knowing Python?

An assistant can help you set it up, but you still need to understand which steps can spend fees, send orders or leave a position open. Use the previews and ask it to explain unfamiliar output before enabling writes.

Should I ask the assistant to trade the five names immediately?

Start with the staged checks. A broad trading prompt skips account, route and borrow decisions that this workflow makes explicit.

Does this require one particular AI product?

No. The prompts describe local file, terminal and scheduling tasks. Your tool's available capabilities and approval controls determine which steps it can perform.

Reference: TradeZero key setup. Continue with additional Alphanume research inputs or the overview.

API documentation checked September 27, 2026. Route availability and account access can change. Examples describe the included starter; they do not establish trading performance.