AlphaForge

AlphaForge is an autonomous trading agent built for verifiable, production-style workflows. Each cycle it pulls real-time market context from the PRISM API (signals, risk, momentum), composes a structured prompt for an LLM, and receives a constrained action: BUY, SELL, or HOLD with reasoning. Before any trade, a dedicated risk module enforces guardrails such as exposure limits, duplicate-position checks, and stop/take-profit style rules. Execution goes through the Kraken CLI (not raw REST), matching the hackathon requirement for an AI-native CLI layer with paper or live modes. Trades, agent runs, and configuration are persisted with Prisma and SQLite for auditability and dashboard views. On the trustless side, the project integrates ERC-8004-style flows on Ethereum Sepolia: agent identity, registration metadata, and EIP-712 trade-intent style attestations tied to the hackathon Risk Router and registry addresses documented in the repository. A Next.js App Router dashboard exposes agent status, signals, and trade history; a separate worker process runs the polling loop. The stack is TypeScript, Next.js, viem, and open-source under MIT. Deployment can use Vercel with cron-triggered ticks; secrets stay in environment variables only. Together, AlphaForge demonstrates transparent behavior, measurable actions, and a path from demo to deployable agent infrastructure.