London · Building the infrastructure coding agents run on
I work on agent-native software — runtimes, registries, gateways, and CLIs that let AI agents do real work in real environments. I write about system design and production architecture at yuanjiwei.com.
- Termio · 430+ ★ · termio.sh — terminal-first development environment for agentic coding: a runtime and multiplexer built for CLI/TUI agents rather than adapted from tmux
- Zypher Agent · 330+ ★ — minimal agent framework with full control over tools, providers, and execution flow
- ACI · 4.8k+ ★ · aci.dev — open-source tool-calling platform: 600+ integrations behind direct function calling or a unified MCP server
- Skilluse · 55+ ★ · skilluse.dev — agent skills registry and CLI: publish, discover, and install skills across Claude Code, Codex, Cursor, and others
- MCP Gateway · 175+ ★ — gateway and control plane governing which tools agents may call, and how those calls are audited
- SigCLI · 290+ ★ · sigcli.ai — authentication CLI and proxy: give agents access, not your credentials
- WeChat Bot · 640+ ★ · wechatbot.dev — WeChat iLink bot SDK for AI agents
- Tunelo · 425+ ★ · tunelo.net — QUIC-based tunnel for exposing localhost, faster than frp, cloudflared, and ngrok
- RunBrowser · runbrowser.com — Chrome extension and CLI that drive your real, logged-in browser over CDP
Also maintaining TTS CLI, speech synthesis from the terminal, and Feedar, an RSS reader built for agents.






