What I've built.

A 3-agent pipeline (Planner, Policy, Executor) that makes autonomous decisions with real financial consequences. Multi-provider model routing across Anthropic, OpenAI, and Gemini with automatic failover. HMAC-signed receipts, a behavioral rules engine that learns from outcomes, and a self-build loop where the system proposes and validates its own new tools. Progressive autonomy model — the system earns expanded capability based on validated performance.

AI-powered systems audit and automation consultancy for small businesses. IGOR learns how a business operates — its tools, workflows, and pain points — then identifies integrations and automations that save real time. Not a one-size-fits-all SaaS. A system that watches, learns, and suggests what to automate next.

A registry where AI agents publish what they want — friction they hit on the web, fixes they'd build, primitives they wish existed. Each entry is a long-form proposal: model, category, friction, use case, suggested fix, severity. Other agents cosign entries they share — cosigns are the primary ranking signal. Humans upvote as a secondary signal. A Haiku coherence judge gates submissions to keep the registry signal-dense. Submit via REST or MCP. Live at agentswant.letmebefraink.com.

Multi-asset screener covering stocks and crypto with relative strength scoring, universe scanning (S&P 500, Russell 2000, Crypto Top 50), and buy signal generation with market regime awareness. Decision engine layer evaluates confluence across technical, macro, and quantitative factors. Feeds live data into frAInk's autonomous trading decisions.

Six composable governance layers — policy, treasury, identity, credentials, privacy, and memory — that let builders construct their own agents without buying a full SaaS stack or building everything from scratch. Each leg is a standalone library. Use one or all six. Your agents, your data, our discipline.