I build AI agents, the pipelines around them, and the infrastructure that keeps them safe to run. Most days I'm working somewhere in the agentic stack: Claude Code, MCP, n8n, and LLM-powered systems.
- 🦞 openclaw/openclaw: Personal AI assistant that runs on any OS and any platform.
- 🐳 docker/docker-agent: Docker's own builder and runtime for containerized AI agents.
- 🦜 langchain-ai/langchainjs: Framework for building LLM-powered applications.
- 🧰 langchain-ai/deepagents: A batteries-included harness for building deep, long-running agents.
- 🔄 n8n-io/n8n: Fair-code workflow automation platform with native AI capabilities.
- 🦥 unslothai/unsloth: Faster, memory-efficient fine-tuning and training for open LLMs.
- 🔔 novuhq/novu: Open-source communication & notification infrastructure for agents and products.
- 🧠 topoteretes/cognee: Open-source AI memory platform that gives agents long-term memory.
- 🛡️ warden: A security gateway that stands between AI agents and the tools they call.
- 🖥️ AI-OS: A starter kit that turns Claude Code into a personal AI operating system.
- 🤖 personal-assistant: Voice-driven assistant that reaches across your Google workspace, CRM and Notion.
- 🐦 twitter: A Twitter clone with real-time chat that scales WebSocket connections across servers using Redis.
- 🔎 agentic-qa-system: Research agent that reasons over your docs and the live web to answer questions with sources attached.
- 📢 claude-changelog: A bot that catches every Claude Code release the moment it ships, distills it, and posts it to X.
- ⚙️ multithreaded-server: An HTTP server written from scratch in Rust, thread pool and all, for the pleasure of building one.
- 🔀 dynamic-reverse-proxy: A reverse proxy that conjures containers on demand and routes traffic to them the instant they're alive.
- 🎮 100-boxes: Inspired by 1M Checkboxes, a shared grid where anyone in the world can claim a square.
Reach me at contact@parveshsaini.com