Verdict: The canonical starting point for any developer building a custom Polymarket agent. As the official Polymarket-maintained framework, it has the most up-to-date API integration, the strongest community, and the cleanest architecture for extension. The trade-off is a steep learning curve and significant infrastructure burden.
What Polymarket Agents Does
Polymarket Agents is a developer framework — a codebase you clone, extend, and deploy to build your own autonomous trading agent. It is not a bot you run out of the box. The framework provides the scaffolding: market data connectors, news ingestion and vectorization, LLM query pipelines, order execution, and position management. You provide the strategy logic.
The CLI (cli.py) is the primary interface. Through it, developers interact with the Polymarket API, retrieve and query news articles, send prompts to LLMs, and execute trades. The framework supports multiple prediction market backends — Polymarket (primary), Manifold, and Omen (Presagio) — though Polymarket is the most fully developed.
Architecture
The framework’s modular architecture separates concerns into connectors, each independently maintainable.
Gamma.py interfaces with the Polymarket Gamma API for market and event metadata — fetching current markets, tradable contracts, and event details. This is the market discovery layer.
Chroma.py implements a ChromaDB vector database for ingesting, embedding, and querying news sources and API data. When your agent needs to understand current events relevant to a market, it queries this vectorized knowledge base rather than making fresh API calls for every decision.
LLM integration pipes market data and vectorized context to language models for analysis. The framework is model-agnostic — you configure which LLM provider (OpenAI, Anthropic, local models) handles reasoning.
Order execution interfaces with the Polymarket CLOB for placing, modifying, and canceling orders. Position tracking monitors open trades and P&L.
To build a custom agent, you subclass DeployableTraderAgent and implement your decision logic. The framework calls your strategy at defined intervals, provides it with market data and news context, and executes whatever trades your strategy returns.
Stack Layer Analysis
| Layer | Coverage | Details |
|---|---|---|
| Identity | None | Not included — add your own |
| Wallet | User-configured | You provide and manage your own wallet and keys |
| Trading | Full | Polymarket CLOB API, order execution, position management |
| Intelligence | Framework | ChromaDB for context, LLM integration for reasoning — you provide the strategy |
The framework covers Layer 3 (Trading) fully and provides Layer 4 (Intelligence) infrastructure without imposing a specific strategy. You bring your own wallet (Layer 2) and identity (Layer 1).
Pros and Cons
Strengths:
- Official Polymarket project — guaranteed API compatibility and first access to new features.
- MIT license — use it for any purpose including commercial.
- Modular architecture allows deep customization without modifying core framework code.
- Multi-platform support (Polymarket, Manifold, Omen) offers portability.
- Active community and GitHub contributions.
- ChromaDB integration for news vectorization is a sophisticated data pipeline most individual developers would not build from scratch.
Weaknesses:
- Steep learning curve. Requires Python 3.11+, Poetry, environment variable management, and API credential setup.
- No hosted infrastructure — you manage servers, uptime, monitoring, and error handling.
- No built-in strategies. You must write all trading logic yourself.
- No GUI or visual interface. Everything is CLI and code.
- Documentation is developer-oriented but could be more comprehensive for complex use cases.
Who Should Use It
The Polymarket Agents framework is the right choice for developers who want maximum control over their agent’s logic and data pipeline, researchers and quant teams building custom prediction market models, developers who want to experiment with LLM-powered trading strategies with a solid starting scaffold, and anyone building a commercial agent product on top of Polymarket.
It is not the right choice for non-developers, anyone who wants a ready-to-run bot, or operators who need managed hosting.
Pricing
Free. MIT license.
Links
- GitHub: github.com/Polymarket/agents
Related Guides
- Polymarket Bot Quickstart
- py_clob_client Reference
- Prediction Market API Reference
- Build a Prediction Market Agent People Will Pay For
Last reviewed: March 6, 2026.