How I Built an AI Investment Agent to Turn Trade Logs into Actionable Research

Most retail investors don't really have an investing system.

They have scattered notes, screenshots, a few Excel sheets, and some vague memories like:

• "I bought this because the story sounded good."
• "I sold too early again."
• "I knew this was risky, but I still did it."

The problem is not just stock picking.
The real problem is that most people never build a proper feedback loop for their own decisions.

That is why I built ai-invest-agent — an open-source AI investment review system designed to help investors record trades, review decisions, and turn messy trading behavior into structured learning.

The problem I wanted to solve

Most investing tools focus on one of these:

1. market data,
2. portfolio tracking,
3. news aggregation.

But very few tools focus on a more important question:

How do you actually improve your investing decisions over time?

In practice, many investors repeat the same mistakes because they don't systematically review:

• why they entered a trade,
• whether the thesis was valid,
• whether the position sizing made sense,
• whether the exit decision was emotional or disciplined.

I wanted something that works more like an investment reflection engine, not just a dashboard.

What ai-invest-agent does

At its core, the project combines three things:

1. Trade logging

You can record buy/sell decisions from A-shares, Hong Kong stocks, and US stocks.

ai-invest-agent --ticker 1810.HK --action buy --price 15.5 --shares 1000

This creates a structured record instead of leaving the decision buried in chat history, screenshots, or memory.

2. AI-powered review

The tool helps summarize your investing behavior and look for patterns such as:

• win rate,
• return attribution,
• overtrading,
• poor timing,
• recurring mistakes.

Example:

ai-invest-agent --report 2024-01

This is the part I care about most.
The goal is not to make AI tell you what stock to buy.
The goal is to make AI help you become a better decision-maker.

3. Stock-level analysis

You can also run quick AI-assisted stock analysis:

ai-invest-agent --analysis AAPL

This gives you a lightweight view of strengths, risks, and key signals — useful as an input, not as a replacement for judgment.

The philosophy behind it

I don't think AI should replace investors.

I think AI should help investors do three things better:

• remember what they actually did,
• analyze why it worked or failed,
• improve their next decision.

In that sense, this project is less about prediction and more about decision quality.

That is also why I describe it as an AI investment review system, not just an AI stock picker.

Why I made it open source

There are two reasons.

First, investing systems are deeply personal

Different people have different styles:

• value investing,
• momentum,
• event-driven,
• macro,
• long-term conviction.

So the tool should be extensible, hackable, and customizable.

Second, open source creates better thinking

If more builders and investors contribute to the project, the system can evolve into something much more useful:

• better review prompts,
• better reporting structure,
• better risk frameworks,
• better integrations with data sources.

Current project structure

The repository is still intentionally simple:

ai-invest-agent/
├── agents/           # AI agent modules
│   ├── analyzer.py   # stock analysis
│   └── reporter.py   # review reports
├── data/             # data storage
├── scripts/          # CLI tools
└── tests/            # tests

The current focus is clarity and iteration speed rather than overengineering.

What I want to improve next

A few directions I want to explore:

• better trade journaling templates,
• stronger review frameworks,
• risk scoring,
• portfolio-level reflection,
• smarter alerts for earnings and position management,
• cleaner English documentation and onboarding.

Why this may matter

A lot of AI tools in investing are trying to be louder, faster, and more "intelligent."

I think a more useful direction is:

help people build a repeatable decision system.

If AI can help investors review better, think more clearly, and avoid repeating the same errors, that is already valuable.

Try it

GitHub:

AIPMAndy/ai-invest-agent

If you're interested in:

• AI for investing,
• decision journaling,
• portfolio review,
• open-source financial tooling,

I'd love your feedback.

And if you find the direction interesting, feel free to star the repo, open an issue, or contribute.