The Problem
Manual crypto trading requires constant market monitoring, thesis generation, risk assessment, and order execution—processes that are time-intensive, emotionally biased, and difficult to scale across multiple assets or venues without significant engineering effort.
What This Does
AutoHedge implements a four-agent pipeline under the autohedge/ package. The Director Agent (autohedge/prompts.py or strategy generation), Quant Agent (autohedge/tools/ for market data analysis), Risk Management Agent (autohedge/workers.py for position sizing), and Execution Agent (autohedge/main.py for order generation) form a structured trading workflow. The tools directory integrates with live data sources via jupiterprice.py, jupitersearch.py, polygonapi.py, and yahooapi.py, while toolsregistry.py coordinates tool selection. The system produces JSON-formatted recommendations and supports full autonomous trading on Solana, as noted in the README reference to Solana support.
How To Use It
Setup: pip install -U autohedge
Configuration: Copy .env.example and populate the required keys. The example file lists: JUPITERAPIKEY — for token price and search tools OPENAIAPIKEY, ANTHROPICAPIKEY — for experimental agents WORKSPACEDIR — default agent workspace directory WALLETPRIVATEKEY — for trading execution
Running it: The entry point is the autohedge CLI command, invoked via autohedge/main.py or autohedge/cli.py. The README's quick start shows: autohedge
Real-World Use
A quant team could instantiate the multi-agent pipeline to automate a Solana-focused market-neutral strategy. By configuring the Director Agent to generate a thesis, the Quant Agent to run technical indicators via jupitersearch.py and polygonapi.py, the Risk Agent to compute position sizes in workers.py, and the Execution Agent to submit orders through the Solana wallet, the system handles end-to-end trade lifecycle with logged outputs. The toolsregistry.py file orchestrates which data source each agent consults, enabling venue-agnostic strategy execution.
Code Health & Issues
Dependencies without lockfile — pyproject.toml and requirements.txt declare dependencies but no lockfile (e.g., uv.lock, requirements-lock.txt) is present, making build reproducibility uncertain across environments. Sensitive data in .env.example — the example file includes WALLETPRIVATE_KEY as a placeholder; committing even a redacted version risks exposing credential-handling patterns. Limited test coverage — 5 test files exist for a project with 51 total files and multiple external API integrations; many code paths (agent pipelines, tool fallbacks) likely lack automated validation. Heavy CI configuration — 24 workflow files in .github/ suggest extensive CI setup, but the low test count relative to file volume raises questions about whether all workflows are actively maintained or if some are vestigial from the parent repository.
The Bottom Line
AutoHedge provides a modular, agent-driven framework for autonomous crypto trading with concrete Solana support and a clear four-stage pipeline. It is well-suited for teams that need structured thesis-to-execution workflows and already have API keys and risk parameters defined. The absence of a lockfile and the presence of sensitive placeholders in the example config are the most immediate practical concerns; these should be addressed before the project is used in production environments.