https://github.com/rnixai/rnix
An operating system for AI agents — Unix philosophy: process, VFS, syscalls. Go 1.26.
https://github.com/rnixai/rnix
agent ai cli go llm mcp orchestration unix
Last synced: 4 days ago
JSON representation
An operating system for AI agents — Unix philosophy: process, VFS, syscalls. Go 1.26.
- Host: GitHub
- URL: https://github.com/rnixai/rnix
- Owner: rnixai
- License: mit
- Created: 2026-03-15T11:20:27.000Z (4 months ago)
- Default Branch: main
- Last Pushed: 2026-06-14T08:07:33.000Z (26 days ago)
- Last Synced: 2026-06-14T10:08:22.167Z (26 days ago)
- Topics: agent, ai, cli, go, llm, mcp, orchestration, unix
- Language: Go
- Homepage: https://rnix.ai
- Size: 10 MB
- Stars: 4
- Watchers: 0
- Forks: 0
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
- Changelog: CHANGELOG.md
- Contributing: CONTRIBUTING.md
- Funding: .github/FUNDING.yml
- License: LICENSE
- Codeowners: .github/CODEOWNERS
- Security: .github/SECURITY.md
Awesome Lists containing this project
README
# Rnix
**The AI-Era Unix — An Operating System for AI Agents**
[](LICENSE) [](https://goreportcard.com/report/github.com/rnixai/rnix) [](https://github.com/rnixai/rnix/stargazers)
[Documentation](https://docs.rnix.ai/) | [中文版](README_zh.md) | [Changelog](CHANGELOG.md)
---
```bash
# You know Unix? You already know Rnix.
rnix -i "Analyze the project structure"
rnix strace 1 # See exactly what your agent did
rnix dashboard # Multi-pane real-time monitoring
rnix gdb 1 # Breakpoints, stepping, context inspection
```
## Design Philosophy
Rnix brings 50 years of proven OS abstractions directly into the world of AI agents. Not a framework. Not a library. A runtime.
**Everything is a Process** — Every agent is a first-class process with its own PID, state machine, and resource table. Use `ps` to list, `kill` to terminate, `strace` to trace — manage agents the same way you manage system processes. Processes can be paused, resumed, suspended, and restarted from checkpoints.
**Everything is a File** — LLMs, filesystem, shell, memory, web search, LSP code intelligence, MCP tools — all unified as VFS devices, accessed through Open/Read/Write/Close. Adding a new capability is just mounting a new device.
**Observability Built In** — Not a logging layer bolted on after the fact. `strace` shows every syscall, `gdb` provides breakpoints and stepping, `dashboard` offers multi-pane monitoring with process tree, timeline, context heatmap, and LLM conversation replay.
**Orchestration is Process Management** — Supervisor trees auto-restart crashed processes. Compose DAGs declare dependencies between agents. Budget Pools manage token allocation. The Intent system decomposes high-level goals into sub-task DAGs. These are OS-level primitives, not application-layer wrappers.
## Quick Start
```bash
# Install
go install github.com/rnixai/rnix/cmd/rnix@latest
# Initialize configuration
rnix init
# Run your first agent
rnix -i "Analyze the project structure"
# Trace syscalls
rnix strace 1
# Real-time dashboard
rnix dashboard
```
## Core Concepts
| Unix | Rnix | Description |
|------|------|-------------|
| `process` | Agent process | Each run gets a PID (UUID v7), state machine, and resource table |
| `/dev/*` | VFS devices | LLM, filesystem, shell, memory, web, LSP, MCP — unified file interface |
| `strace` | `rnix strace` | See every syscall in real time |
| `kill` | `rnix kill` | Terminate any agent from any terminal |
| `top` | `rnix top` | Real-time process monitor with token and time metrics |
| `gdb` | `rnix gdb` | Interactive debugger: breakpoints, stepping, context inspection |
| `ps` | `rnix ps` | List all process states |
| pipe | `rnix compose` | DAG orchestration for multi-agent collaboration |
| suspend/resume | `rnix suspend/resume` | Suspend processes and resume from checkpoint |
## Architecture
```
┌─────────────────────────────────────────────────────────┐
│ CLI & AgentShell │
│ rnix -i / strace / gdb / ps / compose / dashboard │
├─────────────────────────────────────────────────────────┤
│ Agent + Skill Definitions │
│ agent.yaml + instructions.md + SKILL.md │
├─────────────────────────────────────────────────────────┤
│ Microkernel │
│ Process Model / VFS / Context Mgmt / IPC / 50+ Syscalls│
├─────────────────────────────────────────────────────────┤
│ Driver Layer │
│ /dev/llm/* · /dev/fs · /dev/shell · /dev/memory/* │
│ /dev/web · /dev/lsp · /dev/tasks · /dev/tty · /dev/cron│
├─────────────────────────────────────────────────────────┤
│ Host OS + Multi-LLM Providers │
│ Claude · Gemini · Qwen · Anthropic SDK · OpenAI-compat │
└─────────────────────────────────────────────────────────┘
```
## Highlights
- **7 LLM Drivers** — Claude CLI / Cursor CLI / Qwen CLI / Anthropic SDK / Gemini / OpenAI Official / OpenAI-compatible (Ollama, Groq, DeepSeek, etc.)
- **14 VFS Devices** — LLM, filesystem, shell, memory (commit/recall/profile), web, LSP, tasks, TTY, cron, ProcFS, MCP
- **Agent Memory** — Persistent cross-session knowledge with security scanning and async writeback
- **50+ Syscalls** — Stable ABI covering process, context, VFS, IPC, signal, capability, and supervisor
- **Interactive Debugging** — gdb-style debugger + time-travel replay + execution recording
- **Multi-Pane Dashboard** — Process tree, timeline, heatmap, LLM conversation replay, debug mode
- **Token Economy** — Budget pools, SLA contracts, agent reputation scoring, skill synergy matrix
- **Adaptive Security** — Behavioral monitoring, anomaly detection, and auto-suspension
## License
MIT — see [LICENSE](LICENSE).
Full documentation at [docs.rnix.ai](https://docs.rnix.ai/).