By Thomas | financial enthusiast

My markets diary:

May 31, 2026 – Today I discovered Hermes‑Agent, a new GitHub‑based AI tool that lets anyone build a custom research bot without touching a single line of code.

How I stumbled on it

I was scrolling through the #data‑science channel on Reddit when a meme popped up: “Open‑source AI for the masses.” I clicked the link – a GitHub repo titled Hermes‑Agent: Build your own AI research bot. First thought was, “damned, another hype project.” But the README was surprisingly clean. (Works out nicely.)

What Hermes‑Agent actually does

Hermes‑Agent wraps OpenAI’s GPT‑4 in an orchestrator that pulls financial data from APIs like Alpha Vantage, Polygon, and even the SEC’s EDGAR. You just specify the ticker, time‑frame, and the kind of analysis you want – earnings sentiment, valuation multiples, or macro‑factor exposure. The bot then scrapes, cleans, and returns a ready‑to‑publish markdown report.

I didn’t realise how easy it was until I ran a quick test on AAPL. Within minutes, I had a three‑page analysis that matched what I’d normally pull from Bloomberg. The code was hidden behind a simple “build” button; no Python required. (I almost missed this.)

Why this matters in a regulatory climate

AI regulations are tightening—EU’s AI Act, SEC’s guidance on model transparency. Retail platforms are adding AI features, but they’re locked behind proprietary APIs. Hermes‑Agent offers an open‑source alternative that’s auditable and tweakable. I’m excited that I can audit the model’s outputs, tweak the prompt engineering, and even re‑train on my own data if I get the permissions.

Democratizing edge research

Right now, my biggest bottleneck is data ingestion and analysis speed. With Hermes‑Agent, I can spin up a bot for a new IPO overnight and get a sentiment score that feeds into my trade plan. The tool even supports multi‑step reasoning: it can cross‑reference analyst reports, earnings transcripts, and macro data before generating a recommendation.

I’m still learning the nuances—prompt engineering is a craft, and the bot’s default settings are a great starting point but not a silver bullet. Still, the fact that I can iterate on the prompts in a Jupyter notebook, see how the output changes, and share the bot on GitHub is a huge win for us who don’t have a data science team.

Next steps for me

  1. Build a Hermes‑Agent for SPY and compare its sector breakdown to my manual spreadsheet. 2. Experiment with a custom prompt that focuses on ESG metrics. 3. Share my bot on the community forum and gather feedback.

I’m already thinking about how this could affect my trading edge. If I can automate routine research tasks, I’ll have more time to focus on discretionary insights.

Will you give Hermes‑Agent a try and see if it changes how you do equity research?