Your AI quant team, now accepting API keys

What it does ValueCell is a Python-based desktop app that bundles a web interface with a multi-agent system for financial tasks. You configure LLM providers and exchange API credentials, then let agents research assets, monitor news, and execute leveraged crypto trades through Binance, OKX, or Hyperliquid. All secrets stay local on your device; the app talks directly to markets and models without shipping your keys elsewhere.

The interesting bit The project treats “community-driven” literally: it wants to become a decentralized agent marketplace, but right now it is essentially a polished integration layer wiring LangChain/Agno agents to perpetual futures APIs. The guardrails and local-storage pitch matter because you are handing it exchange credentials with real withdrawal risk.

Key highlights

• Supports OpenAI, DeepSeek, Azure, Google, OpenRouter, and other OpenAI-compatible endpoints
• Live trading tested on Binance (USDT-M), OKX, and Hyperliquid; Coinbase, Gate.io, MEXC, and Blockchain.com are partially implemented
• DeepResearch agent scrapes fundamentals; Strategy agent runs automated trades; News Retrieval agent handles scheduled alerts
• Spot trading is implemented as 1× leverage contracts, so the system only works with perpetual futures accounts
• Ships as a downloadable Mac/Windows app or clone-and-run via start.sh / start.ps1

Caveats

• Several exchanges are marked “Partially Tested” and may need debugging; the README itself warns to “prioritize using fully tested exchanges”
• Only contract trading is supported; if you expected spot buys, you will be in perpetuals whether you like it or not
• The roadmap is extensive (securities, European markets, bonds, options, forex) but most of it is aspirational; current reality is crypto futures plus research agents

Verdict Worth a look if you want a local, open-source alternative to closed quant platforms and are comfortable managing exchange API risk yourself. Skip it if you need spot trading, traditional equities, or a battle-tested execution engine — this is early-stage tooling with real money on the line.