For agents
Full-stack engineer building trading infrastructure, internal platforms, and product tooling where reliability, operational clarity, and structured workflows matter.
Currently building trader-facing systems in a production crypto market-making environment, alongside open-source tooling for governed AI-assisted engineering.
Nicolay builds and evaluates AI-native systems, agent workflows, developer tooling, MCP integrations, structured outputs, and emerging browser-agent protocols. This site includes machine-readable metadata and read-only agent-oriented interfaces as an ongoing exploration of human-AI collaboration.
Structor is an open-source AI engineering harness for governing Codex, Claude Code, and other coding agents across repositories. It generates repository-local rules, context routing, task contracts, review standards, and validation scaffolding so AI-assisted engineering becomes a reproducible workflow rather than free-form prompting.
GitHubAI Front Desk turns a small-business website into a source-grounded assistant. It crawls the site, extracts structured business context, generates a reviewable profile, and exposes an embeddable widget that answers customer questions using the business's own content.
Crypto Copilot is a backend prototype for AI-assisted crypto trade planning. It pulls live Bybit market context, builds active-trade state, generates schema-constrained strategy candidates, applies guardrails, and simulates execution without placing live trades.
GitHubThe public MCP demonstration is a read-only JSON-RPC-compatible HTTP endpoint for deterministic portfolio data. It supports initialize, tools/list, and tools/call with read-only tools for profile, projects, experience, and contact data.
POST /api/mcp
{
"jsonrpc": "2.0",
"id": 1,
"method": "tools/call",
"params": { "name": "get_profile", "arguments": {} }
}WebMCP is a W3C Web Machine Learning Community Group draft for exposing browser-native tools to AI agents. It is not a W3C Standard or stable production browser baseline, so this site monitors it without depending on experimental browser APIs.
If WebMCP becomes broadly available behind stable browser APIs and mature security guidance, this site can expose the same public read-only profile and project data through feature-detected browser tools.