Trading bots/agents on Polymarket generated over $60M in profit in 2025-2026. 77% of that came from the Crypto UP/DOWN market - driven by persistent inefficiencies in this segment.
The market keeps evolving: to stay profitable, you need to either continuously hunt for new inefficiencies or build self-learning agents that do it automatically.
That's why Hermes is the ideal foundation for our agent - an open-source, self-learning agent that gained over 100,000 GitHub stars in under 2 months.
Developers are already using it across a wide range of applications, from writing code to building trading algorithms.
By April 27, Hermes Agent GitHub repo surpassed Anthropic's Claude Code in total stars - a clear signal of how useful this framework is for real developers, and exactly why we're choosing it as the foundation for our trading bot.
At the same time Polymarket crypto UP/DOWN market has several inefficiencies, which is exactly how trading bots are classified:
- Arbitrage (Pair Cost) - buys both sides (YES + NO) when their combined price is below $1, locking in a risk-free $0.02–$0.04 profit per pair. Win rate reaches 95–98%
- DCA Bot - waits for one side to drop below $0.35, then averages down until the combined average cost is under $0.99
- Momentum / Latency Bot - monitors BTC spot price on Binance/Coinbase and enters Polymarket during the repricing delay
- Market Maker - places two-sided orders on the 5-minute BTC market, capturing the spread
- AI/ML Bot (Synth SDK) - uses Bittensor AI to forecast probabilities 20 minutes before market close with an edge of 10%+
Examples of successful bots:
• Gabagool22 - legendary arbitrage bot turned $1.2K → $868K on Polymarket
Profile: https://polymarket.com/@gabagool22?via=following
• Sharky6999 - most consistent bot, made $852K PnL with 95.2% win rate.
Profile: https://polymarket.com/@sharky6999?via=following
• @0xe1d6 - fresh HFT crypto-trading bot which made $728K in a month
Profile: @0xe1d6b51521bd4365769199f392f9818661bd907
Today we'll learn how to launch and build the core trading logic for our Hermes Agent on crypto UP/DOWN markets - and then, through live trades and its built-in self-learning capability, let the AI do the heavy lifting for us.
How to set up a Hermes agent in 1 click
Let's shortly discuss what is Hermes agent and why its so good.
Hermes Agent is open-source autonomous agent by NousResearch (an AI research lab funded by Paradigm with $70M), creators of Nomos & Psyche - released on February 25, 2026, with a built-in self-learning loop that makes it more capable the longer it runs.
As I mentioned earlier, it's the fastest-growing framework due to several factors built into its core logic:
- Knowledge Layer: Built-in memory, session search, LLM-Wiki skill, optional Honcho integration. Agent doesn't just answer - it accumulates knowledge over time
- Execution Layer: Multi-agent profiles, child agents, tool system, MCP support, persistent machine access. Agent doesn't just respond - it decomposes tasks, runs them in parallel, and delegates
- Output Layer: Cron jobs, gateway delivery to Telegram/Slack/Discord, Web UI, file outputResults flow back into your real workflow - not trapped in a chat window.
All these factors together make Hermes the brain (not just the hands) of your trading setup - letting it adapt to market conditions rather than blindly following instructions you set once.
Setting up Hermes in 1 click with Atomic
In my previous article on the Hermes weather trading bot, many readers ran into issues installing Hermes via CLI - so I found a more convenient framework for managing your agents.
Atomic Hermes is a native macOS AI assistant - not a browser tab, not a CLI wrapper, not "ChatGPT with buttons". It's an autonomous agent with hands, eyes, memory, and a real workspace.
withе , you can run your agent locally on Mac or deploy it to the cloud using the built-in service feature - linked to your account via email.
Atomic offers 100+ integrations, a large library of pre-installed skills, persistent memory, support for all major AI models (Claude, ChatGPT, Gemini) - and most importantly, one-click setup.
How to set up Hermes Agent with Atomic:
Step 1 - install Atomic app or run in Cloud
Go to https://atomicbot.ai and choose the agent you want to set up on the Atomic main page. In this guide we will use Hermes agent, so I'm choosing this one.
Also, if you don't want to run it locally, you can choose "Run in Cloud" in the right corner of main page → login via Google & get access to same interface.
After app is downloaded, move it into the Applications folder on your Mac.
Step 2 - install agent & connect model API
After we have installed Atomic, we need to plug in a model API for our agent. You can choose to set up free local models (Gemma, Qwen, GLM) or use a paid API from Claude, OpenAI Codex, Google AI, etc.
After the latest OpenAI update, I'm planning to choose Codex as the code engine for my agent logic. All you need to do is buy a ChatGPT Plus plan & connect it to Atomic.
Here we go - we have entered the Atomic dashboard where we can communicate with our Hermes agent and build an up/down trading setup.
You can talk with your agent using the "Chat" tab on the left panel or the CLI if it's more comfortable for you.
Step 3 - connect TG bot to your agent
With Atomic, you can connect your Telegram bot to your Hermes agent in a few clicks.
Go to "Skills" tab in lower left corner → choose messenger Telegram → create bot in Telegram using @BotFather → add bot the API to Atomic
Here we go - a 2-click setup with Atomic and our agent is ready for setting up the trading logic and starting trading on Polymarket.
Setting up Hermes crypto trading logic
Last and most importantly, we need to set up the trading logic for our self-learning crypto trading agent.
Instead of building logic from scratch, I recommend finding GitHub repositories with ready-built logic → feeding this logic to your Hermes agent, and letting it find the most efficient strategy for trading.
List of popular githubs for Polymarket crypto-trading bots:
- JLowo/gengar_polymarket_bot — Quarter-Kelly, Brownian motion probability model, calibrated volatility. Author honest about real-world pitfalls. Circuit breaker, Telegram.
- joicodev/polymarket-bot — Node.js. Black-Scholes, EWMA volatility, Platt recalibration, Brier/Log Loss metrics. Cleanest math.
- aulekator/Polymarket-BTC-15-Minute-Trading-Bot — Production-grade, 7-phase architecture. Grafana, Redis, SL/TP.
- djienne/Polymarket-bot — Two strategies: "Gabagool" (arb) and "Smart Ape" (momentum). Web dashboard, auto-optimization.
- *Up/Down arbitrage (no prediction):*
- Parallax-Trading / Orbital-Alpha — Near-identical Node.js projects. Symmetric ladder, taker arb, on-chain merge to USDC. No Kelly.
With Hermes, I rewrote the logic of those bots so now it uses the latest CLOB2 update by Polymarket, so here is instructions you should send to your Heremes agent:
Step 1: give it *instructions *to build the logic.
As I said before, I'm using an existing GitHub repo to build the logic. What's important to me is that it uses Quarter-Kelly for sizing, which makes it safer.
Step 2: setting up *wallet* for trading
Hermes has built-in safety skills, so after sending the command below, confirm that you understand all the risks and ask him to create wallet he will manage.
Step 3: *executor *updates
Run this prompt so the bot can operate using the latest Polymarket CLOB2
Step 4:* environment/config *setup
Stores all wallet, risk, and live-trading settings in .env for easy setup.
Setp 5: *documentation* set-up
Explains how to install, configure, and safely run the bot in dry-run or live mode.
Step 6: set up *testing* prompt.
Proves the core math and API integration behave correctly and prevents regressions
Step 7: *verification* prompt
Confirms the code compiles and the test suite passes before the bot is considered ready.
And there it is - your Hermes agent is ready to trade on Polymarket up/down markets, powered by custom trading logic, position sizing with the Kelly Criterion, and a self-learning loop that improves with every trade.
*note:* i recommend starting with small $1-$2 trades to let your Hermes agent learn, and build its own trading logic based on executed trades. Let the magic of AI build its own strategy.
Conclusion:
Polymarket trading bots have already taken a large share of the profit pie from manual traders, and this percentage keeps increasing daily.
Nowadays, with agentic frameworks like Hermes and @atomicbot_ai, you don’t need to be a senior-level developer to build your own agent.
All you need is a good ML model, right prompts and time for training.
