"6 free GitHub bots for trading on Polymarket." That's the pitch from a thread by the trader @recogard, and it's easy to see why it spread. Each of the six projects comes with a one-line promise: backtest any idea, mirror wallets that already work, farm liquidity rewards, or hand the whole thing to Claude.
The timing makes sense. Prediction markets have grown into a real venue, and the same agent tooling that now writes code and drives a browser can also watch an order book, react to news, and place an order. A lot of that tooling is genuinely open source and free to clone.
But a thread is not a review, and "free" doesn't mean "safe." So we cloned and read all six repositories. Below is what each one actually does, how far along it is, and what the thread skips.
Before anything else: this is not financial advice. These are programs that can move real money on a volatile, and in some places restricted, market. Start with dry-run modes, small amounts, and the assumption that any "strategy" is a hypothesis rather than income.
The six tools, sorted by the job they do
Automating trades isn't one task—it's several. You research an edge, you test it against history, you execute, and some people add liquidity to earn rewards on top. The six projects line up surprisingly well along that path:
| Project | Language | Stars* | License | Job |
|---|---|---|---|---|
| prediction-market-backtesting | Python / Rust | ~1.2k | Mixed (MIT + LGPL-3.0) | Backtest strategies on historical markets |
| AutoPilot | TypeScript | ~45 | MIT | Self-hosted AI workspace to research and trade |
| Prediction-Markets-Trading-Bot-Toolkits | Rust | ~480 | MIT | Multi-strategy execution engine |
| polymarket_lp_tool | Python / Rust | ~570 | None | Manage limit orders for liquidity rewards |
| PolyWeather | Python | ~320 | AGPL-3.0 | Weather-market analysis and signals |
| polymarket-mcp-server | Python | ~700 | MIT | Give Claude tools to trade and monitor |
*Star counts are rounded from GitHub and change constantly; they're a rough popularity signal, not a quality score.
1. Test an idea before you fund it: prediction-market-backtesting
The most-starred project here doesn't trade anything. It's a backtesting framework for prediction-market strategies, built on top of NautilusTrader with custom exchange adapters, currently focused on Polymarket.
You feed it a strategy and historical market data, and it replays the book to show how the strategy would have performed—P&L, win rate, drawdown, Sharpe, and a cumulative Brier score, which is a more honest metric for probability predictions than plain returns. The charting is the standout feature.
This is the project to reach for if you have an idea like "buy YES below 30 cents on political markets" and want to know whether it survives contact with real data. It needs Python 3.12+ and a working knowledge of NautilusTrader's model. The license is mixed—MIT for the author's code plus LGPL-3.0 for the NautilusTrader parts—so check the NOTICE file if you plan to build on it commercially.
⚠️ Actively developed: the repo describes live/sandbox plumbing for fast-moving BTC markets as alpha. Backtesting, not live trading, is its strength.
2. An all-in-one workspace: AutoPilot
AutoPilot is the opposite approach: instead of one focused tool, it's a self-hosted AI workspace that bundles research, execution, and portfolio tracking behind a local dashboard and a chat interface. You install it with npm, run autopilot onboard, and open a web UI on your own machine.
It claims a lot: 118+ strategies and presets, market data across Polymarket, Kalshi, Binance, and several DeFi venues, plus risk controls like position limits, a daily-loss cap, and a kill switch. There's an MCP server and support for a dozen messaging channels, so you can talk to it from Telegram or Discord.
The obvious caveat is that "does a lot" and "does any of it well" aren't the same. It's the smallest project on this list by stars, and it connects to real accounts and can submit real transactions. The README itself recommends starting in dry-run mode, and that's the right call. Treat it as a tinkerer's cockpit, not a finished product.
3. A shared engine for many strategies: Prediction-Markets-Trading-Bot-Toolkits
Written in Rust, this one is built around a single idea: one execution engine and one risk layer, with venues added as adapters. The README maps out ten strategies—copy trading, cross-market arbitrage, spread farming, an order-book-imbalance signal, and more.
Here's where the original thread oversells it. The tweet lists copy trading, arbitrage, market making, whale alerts, and spread farming as if they were all ready. The repository's own description is more careful: copy trading is production-ready, and the other nine strategies are still in development on the same core. The author is also explicit that the hosted service is closed, and that running it yourself is the supported path.
On the plus side, it ships prebuilt binaries for macOS, Windows, and Linux, so you don't need a Rust toolchain, and it runs in dry-run by default (enable_trading: false) until you deliberately flip it.
4. Managing limit orders for rewards: polymarket_lp_tool
This project targets a specific, less glamorous edge: liquidity rewards. Polymarket pays traders to keep resting limit orders near the midpoint, and this tool watches those orders and repositions them as the order book moves.
The important detail is what it doesn't do. The README is blunt—"this is not an automatic market-making bot." It never places your opening orders; you place them by hand, and the tool only manages the ones already on your account, deciding to keep, cancel, or re-quote them. There's a main Python program plus a newer Rust rewrite that the author labels experimental.
Two things to note before you rely on it. First, it has no license file, which by default means all rights reserved—so read the terms rather than assuming you can reuse the code. Second, the README includes a referral link for the author; useful, but a reminder to evaluate the tool on its merits.
5. A specialist for weather markets: PolyWeather
Weather is one of the more tractable corners of prediction markets because it has real underlying data—airport temperature readings, forecasts, and official observations. PolyWeather aggregates those sources across dozens of cities and turns them into calibrated probability estimates and market signals, then surfaces it all through a dashboard and a Telegram bot.
The engineering is serious: it blends multiple weather models, tracks settlement stations city by city, and flags where its model probability diverges from the market price.
The catch is licensing and scope. The repo moved to AGPL-3.0, and the README is direct that the public code is a subset: private production data, risk rules, position-sizing logic, and the actual trade-execution code are not included. So it's best read as a research and signal layer, not a turnkey trading bot.
6. Letting Claude do the clicking: polymarket-mcp-server
The last project takes the MCP route. It's a server that exposes Polymarket to Claude as a set of 45 tools—search markets, read the order book, place and cancel orders, track positions, and stream live prices—so an AI assistant can research and execute through conversation.
It's one of the more polished repos here, with rate limiting, EIP-712 order signing, and safety controls like order-size limits and a confirmation step for large trades. There's also a web dashboard.
The trade-off is the one inherent to the whole category: to trade, you supply a wallet private key. That key is your funds. Run it in a controlled environment, watch what the model actually does, and never hand an unattended agent more authority than you'd give a stranger.
What the thread leaves out
Three things are worth repeating after reading all six repos:
- "Free" describes the code, not the risk. Most are MIT-licensed, but one has no license at all, and another is AGPL with a closed-source core. If you plan to build on any of them, check the terms first.
- Maturity varies wildly. One project has over a thousand stars and years of development; another is a few months old with a few dozen. None of the "strategy" descriptions in the thread are guarantees—several of these tools explicitly warn that their strategies are unproven or experimental.
- You are the custodian. These are self-hosted tools that connect to your wallet and your API keys. Start in dry-run, keep position sizes small, and assume the market is faster and better funded than your script.
Who this is for
If you write code, understand how a limit order book works, and are comfortable running a server, this stack is genuinely interesting—especially the backtesting framework and the MCP server, which are useful even if you never trade a cent. If you're looking for a button that turns prediction markets into passive income, none of these six is it, and the ones that promise the closest to that are also the ones asking for your private key.
