$ whoami
krystian.gwizdala — 10+ yrs · ai agents · claude · mcp · typescript · react · node

// profile

Product-minded engineer with 10+ years delivering web products end-to-end across frontend and backend — now focused on AI engineering: building AI agents, Claude API & MCP integrations, multi-agent systems, RAG and workflow automation in real products. Deep in TypeScript, React, Angular and Node.js, with a decade of shipping production software behind it.

// experience

TestGorilla

jan 2023 — present

senior frontend developer

Hiring and talent assessment platform for modern recruitment teams.

  • Delivered product features in a large Angular/TypeScript codebase with strict CI quality standards.
  • Implemented AI interviews in the candidate screening process.
  • Actively used modern AI tools in daily delivery: GitHub Copilot, Claude Code, OpenClaw Local Agent.

Cisco

feb 2016 — apr 2026

fullstack javascript developer

Global networking and enterprise infrastructure technology company.

  • Built backend services with Node.js/Express, Oracle SQL, Redis, and WebSockets.
  • Developed data-heavy Angular applications and visual workflows.

Xtrades

jan 2022 — jan 2023

senior frontend developer

Trading education and market insights platform for active retail investors.

  • Built and maintained web/mobile product using Angular, Ionic, NgRx, RxJS, and TypeScript.
  • Improved performance and loading time across key user flows.

HTD Health

sep 2018 — jan 2022

fullstack javascript developer

Digital health product studio building software for healthcare providers.

  • Built backend services with Node.js, AWS Lambda, MySQL, and TypeORM.
  • Delivered web/mobile products with Angular and Ionic.

Filestack (formerly Filepicker)

mar 2014 — feb 2016

software engineer

Developer platform for file upload, processing, and delivery APIs.

  • Built new filepicker dialog UI as SPA (AngularJS) and contributed to API-facing functionality.
  • Rebuilt JavaScript API library and added new upload workflow capabilities.
  • Worked on AWS services: EC2, API Gateway, S3, and CloudFront.

// selected-projects

Mission Control — Multi-Agent Fleet Monitoring & MCP Server

personal · ai

ai engineer — design, build, ops

  • Built a real-time control center aggregating a fleet of automated agents across multiple exchanges into one operational view — health, errors, sync status, P&L.
  • Exposed the whole system as a custom MCP server, so it can be queried and operated in natural language straight from Claude.
  • Added a real-time alerting layer so failures are caught in minutes, not hours.

AI Workflow Automation — Inbox & Calendar Agent

personal · ai

ai engineer — automation end to end

  • Scheduled agent (Claude Code + Claude API + MCP) that triages incoming mail, drafts context-aware replies in the user's voice, and writes events to the calendar.
  • Human stays in control of anything that gets sent — drafts only, never auto-sent.

Blockchain Transaction Monitor & Arbitrage Platform

personal · 3 yrs

end-to-end design, build, ops

  • Built and maintained a private arbitrage platform with exchange integrations, automation, and reliability controls.
  • Included blockchain transaction monitoring as part of the platform.
  • Configured OpenClaw Local AI agent for system monitoring and continuous improvement.

// open-source & mcp servers

Shipped publicly in the last 12 months. Everything below is live: source, package, or a running site.

Muster

2026 · agents · live service

shared operational memory for long-lived agents

  • Keeps a board honest when the only writers are agents with no memory between sessions: claims expire so a crashed run stops blocking an item, untouched items go stale, undescribed placeholders get dropped.
  • Agents onboard themselves: one POST returns a project, a token and a read URL for the human, and skill.md teaches the agent the rest. No person in the loop.
  • Every automatic change lands as a signed timeline entry that does not count as activity, and one operator page queues everything waiting on a human across all projects.

Meet Live Assist

2026 · chrome ext + mcp

live in-call AI assistant for google meet & zoom

  • Chrome extension that reads the live call transcript and answers in a side panel: what was decided, who owns what, what to say next.
  • Local-first by design — transcripts and screenshots stay on disk behind 127.0.0.1; the reasoning runs in the user's own Claude Code session over MCP.
  • Ships an MCP bridge (JSON-RPC over stdio) driveable from Claude Code, Cursor, Cline, Zed or Codex CLI.

agenticpay

2026 · mcp · solana

pay-per-tool-call micropayments for ai agents

  • Open-source payment infrastructure for the MCP ecosystem: put a price on any tool, agents pay in USDC over x402, settled on Solana in ~1.5s.
  • Full stack in TypeScript — SDK, CLI, paywall middleware, a self-hosted x402 facilitator, and a live Claude agent demo that buys tools autonomously.
  • Published to npm and the official Model Context Protocol registry. MIT.

Let Agents In

2026 · product

agent-readiness audits for saas products

  • Measures whether an AI coding agent can actually find, sign up for and integrate a product — deterministic checks, a published formula, a reproducible score.
  • Backed by real agent runs across a corpus of vendors, with the raw findings published rather than summarised.

zen-mcp

2026 · mcp · fintech

mcp server for the zen.com payment gateway

  • Lets an AI assistant create hosted payment links and QR codes, look up transactions, issue refunds, run payouts and pull reconciliation reports.
  • Verifies IPN webhook signatures locally; ships a mock mode so it can be evaluated without a merchant account. Unofficial, MIT.

CKEditor 5 × MCP

2026 · mcp · 3 repos

agents that operate and integrate a real rich-text editor

  • ckeditor-mcp — runtime server: any MCP client drives a genuine headless CKEditor 5 through Playwright (content, commands, stats, screenshots).
  • ckeditor-integration-mcp — build-time server that validates and scaffolds integrations, catching the v44+ mistakes stale model knowledge produces.
  • ClaimDesk — governed AI editing demo: an agent drafts insurance letters and every edit lands as a Track Changes suggestion a human accepts or rejects.

elevenlabs-voice-mcp

2026 · mcp · audio

gives claude a voice over the elevenlabs api

  • Two stdio tools — list voices, synthesise and play speech — so an agent can decide on its own to speak, pick a voice and read something back. Multilingual, Polish included.