Agentic engineer building the infrastructure layer for AI agents: search and retrieval, memory systems, MCP tooling, messaging integrations. I turn APIs and workflows into reliable agent tools. The engineering is in the loop around the agent, and the same patterns port across OpenClaw, Hermes, Nanoclaw, and standalone MCP/PyPI packages.
19 merged PRs to OpenClaw core · 74k+ ClawHub registry downloads across 12 skills · flagship web-search-plus at 382★
Latest: OpenAgentFleet, a local-first macOS runtime for AI agents with isolated computers and explicit approvals.
🌐 robbyczesany.com · 📫 Open to agent-infrastructure work: robby@robbyczesany.com
OpenAgentFleet is a local-first macOS app for running AI agents with explicit control. Grok Build, Codex App Server, and OpenCode in one workspace. Browser and desktop tasks run on an isolated Linux computer you can watch, stop, or take over. Approvals go through the controller, data stays local, and you can steer it from your phone over Tailscale.
Go controller · Tauri + React shell · Apache-2.0 · signed and notarized alpha for Apple Silicon → openagentfleet.xyz
web-search-plus is a multi-provider search engine for AI agents. 13+ search providers, 6+ extraction providers, auto-routing, opt-in research mode, multi-LLM fallback chains.
One engine, three deployments: hermes-web-search-plus (382★), web-search-plus-plugin for OpenClaw, and a standalone MCP server (web-search-plus-mcp on PyPI).
- lucid-dreamer — nightly memory reasoning; cleans agent memory while you sleep
- skillminer — scans agent memory, detects patterns, drafts new skills for human review
- topic-monitor — scheduled topic monitoring with AI importance scoring, included in DigitalOcean's OpenClaw Skills guide as proactive-research
Also built: agent-chronicle (AI-perspective diary), roundtable (multi-agent debate council), and OpenCami (a web client for OpenClaw).
I design the system, let agents accelerate implementation and research, review every commit by hand, test the behavior, and ship the parts that hold up. Human decisions, agent-assisted execution, tight feedback loop.
- OpenClaw contributor. 19 merged PRs: Telegram, Discord, session management, formatting, bug fixes
- 74k+ ClawHub registry downloads across 12 published skills
- Skills listed in awesome-openclaw-skills and awesome-hermes-agent
Go (local agent runtime) · Python (engines, MCP servers) · TypeScript (plugins, web) · React + React Native · Tauri · Node.js · Vite · Expo · SQLite · Docker/Colima sandboxing · multi-LLM orchestration with fallback chains
🌐 robbyczesany.com · 📫 robby@robbyczesany.com · openagentfleet.xyz · websearchplus.xyz · Graz, Austria (CET)




