Linggen is an open-source agent system where agents, skills, and scheduled missions are all markdown files. Drop a .md file, get a new agent.

  - Multi-agent — agents delegate to each other, run concurrently. Message a running agent mid-task and it
  adapts (cooperative interruption)
  - Any model — Ollama, OpenAI, Claude, Bedrock, any OpenAI-compatible endpoint. Route per agent with
  auto-fallback
  - Skills as files — portable Markdown skills, compatible with Claude Code and Codex (Agent Skills
  standard)
  - Scheduled missions — cron-based autonomous tasks without human-in-the-loop
  - Web UI + TUI + VS Code — all share real-time session state via SSE

  Rust backend (axum, tokio, ratatui). React 19 frontend embedded via rust-embed. Native tool calling for
  all major providers. Workspace-scoped file ops with per-agent permissions.

  Early stage. The name 灵根 (Líng Gēn) means "spiritual root" in cultivation fiction — the innate
  foundation that determines what you can become.



  GitHub: https://github.com/linggen/linggen

  curl -sSL https://linggen.dev/install.sh | bash && ling

  Author here. Built this because I wanted one system where I could drop a markdown file and get a new agent
   — for coding, but also for scheduled code reviews, architecture checks, or anything else.

  The file-based approach is the core idea. An agent is 15 lines of YAML + markdown. A skill is a SKILL.md
  directory. A mission is a cron entry pointing at an agent. No SDKs, no plugins, no code changes — just
  files.

  Happy to go deep on the Rust runtime, multi-agent delegation, or anything else.