It's Adam, also known as overment. I build AI agents, the tools they use, and products around them:
- Alice — a desktop AI app and my sandbox for exploring what generative AI can and cannot do, in development since the "ChatGPT moment" of late 2022.
- MEGA — architecting a 4-week online program for developers building with AI: its curriculum, lesson framework, practice assets and tooling, and editorial publishing.
- easy.tools — checkout, delivery, and automation for creators selling digital products; built with a friend for ourselves, until our clients wanted it too.
- eduweb.pl — an educational platform with tech and design courses, 15+ years, 200,000+ clients. Education sits at the core of my mission.
I write a lot and share what I learn on X and YouTube. Everything else lands in my Wonderlands-powered digital garden at overment.ai.
wonderlands — my biggest project: a self-hosted, multi-tenant home for conversational AI agents and my digital garden. Agents run through a background-job-based multiagent runtime, letting them take multiple turns, call MCP tools, drive a browser sandbox, run on schedules, and wait on input without blocking anything. The same platform publishes my workspace vault as overment.ai: notes and essays on concepts, tools, books, mental models, and ongoing projects.
Everything my agents touch, they touch through the Model Context Protocol. All TypeScript + Hono; HTTP servers use the Streamable HTTP transport.
| Server | What it does |
|---|---|
| streamable-mcp-server-template | The base for all of them — dual runtime (Node.js and Cloudflare Workers from one codebase), five auth strategies including OAuth 2.1, encrypted token storage |
| Spotify | Search music, control playback, manage playlists and saved songs |
| Linear | Manage issues, projects, teams, cycles, and comments |
| Google Calendar | Manage events, check availability, schedule meetings |
| Gmail | Search threads, read messages, manage drafts, organize the inbox |
| Google Maps | Search places, get details, plan routes — built for location-aware agents on mobile |
| Tesla | Vehicle control via the Tessie API |
| Replicate | Image generation and editing with official models: Flux, SDXL, Seedream |
| Resend | Email and newsletter management |
| Firecrawl | Web scraping and search with batch support and rate limiting |
| ElevenLabs | Text-to-speech, speech-to-text, voice management |
| YouTube | Search videos, explore channels, browse playlists |
| Files | Sandboxed file access with checksum-safe edits (stdio) |
| Video | Video understanding via Gemini — YouTube and local files (stdio) |
voice-to-notes — turns voice recordings (e.g. Superwhisper) into daily Markdown notes with Gemini: classifies each recording into configurable categories, runs per-category prompts, and keeps a rolling memory per category. Ships as the spoken-English coach I use every day.