I kept hearing about Kalshi. The commercials, the mentions, and then one morning CNN was talking about Kalshi prediction odds like they were a weather report. So I went and looked. And I had no idea what I was seeing.
Kalshi is really hard to understand when you're new to it. You get markets, contracts, prices that are also probabilities, volume, movement, and none of it tells you what's actually worth paying attention to. I wanted something that would translate what I was looking at into something I could understand and, ideally, act on. That was the whole original goal: make the firehose legible.
Then, of course, I kept adding to it, because once you can read the flow you start wanting an edge in it. Who doesn't. So Trade Hunter grew from a translation layer into a translation layer with detection on top. This is a writeup of what it became, why I made the design choices I made, and the parts I'm still not sure about. I'd rather you poke holes in it now than find the breakpoints the hard way. If you think a decision here is wrong, please let me know. The comments are the point of this post, not an afterthought.
Where this came from
Basically, I couldn't read Kalshi, and I wanted to. Trade Hunter is the original tool I built to fix that for myself, and it's the one I still run when I want a live view. So this isn't a polished sequel to anything. It's the thing I built because I wanted it to exist, flaws and design bets included, and I'd rather show you those directly.
The core idea never changed even as I piled features on: watch live Kalshi WebSocket feeds and surface an unusual move while it's still moving, rather than reading about it after the fact, or on CNN the next morning.
What it actually does
Trade Hunter subscribes to live Kalshi feeds across every market you track. Multi-contract series like the Fed rate decisions or who will win Top Chef fan out automatically to all open contracts, so you point it at one thing and it watches the whole family. When something moves unusually, it runs the move through a spike detector with configurable volume, price, baseline, and cooldown thresholds, and surfaces it on a local dashboard at 127.0.0.1:8765 with a signal feed, trade flow, and a market tape.
The signals sort into tiers, roughly from "noticeable" to "this one cleared every gate at once." The tier I care about most is the top one, where the score, price, and volume conditions all trip simultaneously rather than one at a time.
That's the boring part. Here's the part I actually want feedback on.
The two-layer read
A raw spike detector produces a lot of noise. Volume jumps for dull reasons constantly. So there are two AI passes sitting on top of the detector.
Layer one is a per-signal analyst. Every spike gets an immediate read: signal, noise, or uncertain, with a direction, a confidence level, and a plain-English rationale for why the flow looks the way it does. This runs on Claude by default with automatic fallback to OpenAI/ChatGPT, Gemini, Perplexity or XAI/Grok. You bring your own API keys. If you configure neither, the detector still runs, you just don't get the narrated readout.
Layer two is a tuning advisor. This is the piece I'm most attached to and least certain about. It looks across your recent analyst-labeled signals, finds patterns in your false positives, and suggests concrete threshold changes you can apply with one click. The claim I'm careful to make honestly: the underlying model doesn't retrain. Nothing about the LLM gets smarter. What improves is your detector configuration, tuned to the specific markets you watch, using your own accumulated signal history. It compounds in the sense that a spreadsheet of good heuristics compounds, not in the sense that a neural net does. I'd rather undersell that than dress it up.
The one detector rule I'm proud of
Whale clustering. A single large trade isn't that interesting; large trades happen. What's interesting is several of them bunched together in a way that's unlikely to be coincidence. So the "whale-cluster" flag fires when at least three trades at the 99th-percentile volume land inside a 120-second window, where a Poisson model puts the probability of that clustering by chance below one percent. That's the closest thing here to "coordinated flow" that I felt comfortable labeling as such. Tell me if the threshold is naive.
Run it and walk away
Here's the feature I probably undersell: Trade Hunter pushes signals to Discord over webhooks, with per-topic channel routing so different markets land in different channels. That changes how you actually use it. You don't park yourself in front of the dashboard. You start it, let it run, and the analyst readouts come to you: the signal, the tier, and the plain-English read, dropped into Discord wherever you are. Then you go act on Kalshi from your phone.
For a tool whose entire job is catching a move while it's still moving, being able to leave your desk and still not miss it is most of the point. The dashboard is where I look when I'm at the machine. Discord is why I don't have to be.
The decisions you're going to question
I'd rather raise these myself than have them raised for me.
It's self-hosted, and that's deliberate. Every other Kalshi monitor I found is a hosted SaaS. Your data lives on their server, their model improves for their benefit, and when you stop paying you're left with nothing. Trade Hunter runs on your machine, persists to a local SQLite database with a seven-day retention window, and the source is yours to read and modify. I work in information security by day, and "your market positions and your API credentials leave your machine and sit on someone else's box" is a tradeoff I didn't want to make or ask anyone else to make.
Live mode touches your Kalshi private key, and you should be suspicious of that. This is the honest trust ask. To go live you drop a .env file and your Kalshi private key next to the program, and it signs its own feed subscriptions. That's a real thing to hand to software from a stranger on the internet. My answer isn't "trust me." My answer is that the full Python source ships with it, the key never leaves your machine, and there's no server for it to phone home to. If that's not enough for you, that's a completely reasonable position, and I'd rather you inspect the source than take my word.
There's an exe, and there's also full source. The exe is there so someone who wants signals but doesn't write Python can double-click and go, though that convenience is Windows-only. The source is there so someone who does write Python isn't stuck with my choices, and it runs anywhere Python 3.11+ does: Windows, Mac, or Linux, with a little extra setup on Mac and Linux. If shipping a binary at all feels wrong to you for a tool that handles credentials, I want to hear that argument, because I went back and forth on it.
What I'm deliberately not claiming
It doesn't place trades. It's informational only. It surfaces things for your review and you decide what to do on Kalshi. It's not financial advice, it doesn't promise outcomes, and the addressable audience is frankly narrow: people who trade on Kalshi and are comfortable running local software. I'm not going to pretend this is for everyone.
Try it without trusting me at all
Simulation mode is the whole point of this section. It runs the entire dashboard on synthetic data with no credentials, no API keys, and no Kalshi account. You can see exactly what the detection, the tiers, and the two-layer read look like in about sixty seconds, and decide for yourself whether the thing is any good before money or trust enters the picture.
What I would like from you
Where did I get it wrong? The Poisson threshold, the "it doesn't retrain" honesty, the binary-plus-credentials decision, the whole self-hosted premise, or something else? I built this because I wanted it to exist and I use it myself, but I've been staring at it alone for too long. Is there something that bothers you about this? Tell me about it.
If you run the sim version and decide you want it pointed at real Kalshi market flows, that's the commercial build: same dashboard, live feed, full Python source, your keys, all running on your machine. It's on an introductory launch price right now at tradehunter.site.
Top comments (0)