Senior Software Engineer · HCMC

Phạm Văn Quý

Self-hosted AI agents, MCP toolchains, HTTP tunnels, and automation pipelines — built to run on your machine.

Own your stack

Agent runtimes, MCP toolchains, and render pipelines that run on your iron — not a rented SaaS middleman.

11+ years shipping NestJS, React/TypeScript, and Kubernetes in production. Day-to-day focus: multi-agent orchestration, tool calling, streaming UIs, and wiring LLMs into real product workflows.

Open source

Three self-hosted projects — agents, tunnels, and video rendering. Code on github.com/phamvanquyit.

TypeScript · Bun

Docker one-liner

raw-agents

The Brain — self-hosted AI Agent Manager

Run multiple AI agents on your machine — web UI, realtime chat, cron schedules, and custom tools. No cloud SaaS dependency.

  • Multi-provider: OpenAI, Anthropic, Gemini, OpenRouter via LangChain
  • MCP servers (SSE / Streamable HTTP) + custom tools in JS/Python
  • Builtin stealth browser: navigate, click, fill, snapshot, screenshot
AI AgentsMCPBunHonoReactSQLite
View on GitHub

Rust

Traefik · TLS · yamux

ZoTunnel

Your VPS, ngrok-style tunnel

Expose a local app through your own VPS — subdomain routing, automatic Traefik TLS, ngrok-style CLI. Traffic stays on your machine, not a middleman SaaS.

  • Server + client in Rust, multiplexing over yamux
  • Traefik dynamic config: connect writes the router, disconnect removes the file
  • CLI zotunnel: http 3000, custom subdomain, background mode / upgrades
RustTraefikTLSCLINetworking
View on GitHub

TypeScript · FFmpeg

npm package · Skia

qpv-json2video

JSON in → MP4 out

Render video from JSON with Skia canvas + FFmpeg. Multi-track timeline, karaoke captions, full effects — built for automation and AI content pipelines.

  • 8 element types: Text, Image, Video, Caption, Shape, SVG, Waveform, Timer
  • 17 animations + 11 scene transitions, chroma key, mask, blend modes
  • Caption word-level highlight & CapCut-style word-by-word reveal
VideoFFmpegSkiaCanvasnpm
View on GitHub

Build with agents

Looking to collaborate or need consulting on agent architectures — orchestration, MCP tools, self-hosted runtimes, or shipping LLM features into production? Reach out.