Ragon ## Private AI on your DGX Spark. One command. **A fully local alternative to cloud coding agents.** A coding agent, a chat assistant and a desktop app, running entirely on your own GB10. No cloud. No account. No API bill. No telemetry. [**Install**](#install) · [**Try it**](#try-these-first) · [**Proof**](#-your-code-never-leaves-your-desk) · [**Docs**](docs/getting-started.md) · [**Releases**](https://github.com/dreamference/ragon/releases) [![Latest release](https://img.shields.io/github/v/release/dreamference/ragon?label=release&color=FF6B35)](https://github.com/dreamference/ragon/releases/latest) [![Telemetry: none](https://img.shields.io/badge/telemetry-none-242A32)](docs/privacy.md) [![Runs on any GB10](https://img.shields.io/badge/runs%20on-any%20GB10%20%C2%B7%20128%20GB-555555)](#runs-on-every-gb10) [![GitHub stars](https://img.shields.io/github/stars/dreamference/ragon?style=flat&color=FFB400)](https://github.com/dreamference/ragon/stargazers) [![License: AGPL v3](https://img.shields.io/badge/license-AGPL%20v3-blue)](LICENSE) | **0** | **$0** | **262K** | |:---:|:---:|:---:| | phone-home connections, verified by `audit egress` | per token, forever | tokens of context |
## RBRB: the Ragon-Builds-Ragon benchmark ![RBRB latest share](images/rbrb-badge.svg) RBRB measures, per ISO week, the share of development tokens generated by Ragon (the product's own local model) versus by any other model, over all work on this repository and its engine. Every landed commit carries a `Ragon-Tokens: ` trailer, and the primary measure is generated (completion) tokens on both sides. Ragon's work passes the same tests and review gates as everyone else's. The weekly target tracks the measured share (last week's share minus 2 points, never below a 2 % floor). Quality first, progress second: the share never rises at the cost of quality, and a regression traced to Ragon-made work freezes the increase until it is fixed. The first point is week 2026-W46, complete and counted from the gate's own usage counters (generated tokens only; the hand-measured partial week stays in the data as a note); the goal is a curve that rises over years without lowering quality. Data: [docs/rbrb.json](docs/rbrb.json). ![RBRB](images/rbrb.svg) Latest value: **8.00 %** (2026-W46, closed); **9.84 %** so far in 2026-W47 (the running week). ## Install ```bash curl -fsSL https://github.com/dreamference/ragon/releases/latest/download/install.sh | bash ``` On a GB10 you can start it and leave: it asks for your sudo password once, at the start (and whether to install pending NVIDIA system updates), then sets up the host, offers the machine to your local network, downloads the model (~20 GB) and starts serving it, and ends with a summary of every step. Then point Ragon at a project: ```bash cd ~/your-project && ragon # that's it ``` Works on the **NVIDIA DGX Spark** and every GB10 machine from Acer, ASUS, Dell, Gigabyte, HP, Lenovo and MSI. The script verifies every download against the release's checksums and prints each `sudo` command before running it. [Read it first](install.sh) if you like. **On your laptop:** run the same one-liner on a Linux laptop (Intel/AMD or Arm) or a Mac (Apple silicon or Intel). It installs the client, which uses the model on your GB10 over your network. It also puts the desktop app in Applications on a Mac (`--no-app` leaves it out). On Windows 11, in PowerShell (a preview; it installs the desktop app too): ```powershell irm https://github.com/dreamference/ragon/releases/latest/download/install.ps1 | iex ``` `ragon update` then updates the terminal agent and the app together. Other ways to install: the release page's dmg and Windows installer are unsigned previews that macOS and Windows stop the first time ([first-launch steps](docs/desktop.md#other-ways-to-install)). Already using a coding agent? Paste this into it: ```text Install Ragon on this GB10 from https://github.com/dreamference/ragon (one-line installer in the README), then run ragon-admin server start. ``` ## Try these first ```bash ragon "explain this repository to me like I'm new on the team" ragon "the tests are failing: find out why and fix them" ragon "add input validation to the signup form, with tests" ``` Or queue work for tonight, inside a session: ```text /night add upgrade every dependency and fix whatever breaks ``` Ragon works through the queue while you sleep, each task on its own git branch. You review the branches over coffee. ## Why switch | | Cloud coding agents | Ragon | |---|---|---| | Where your code goes | Their servers | Nowhere | | Account needed | Yes | No | | Cost per token | Metered | Zero | | Rate limits | Yes | No | | Works with the network unplugged | No | Yes | | You can verify what leaves | No | `ragon-admin audit egress` | --- ## Why Ragon ### 🔒 Your code never leaves your desk Cloud coding agents send your source code to someone else's servers. Ragon sends it nowhere: the model runs on your GB10, and nothing in Ragon phones home. No usage analytics, no metrics, no sign-in, no "anonymous" crash reports. That makes it usable for code you're not allowed to upload: client work, regulated data, unreleased products. **And you can prove it.** One command runs a real Ragon session under `strace` and lists every network connection it made: ```console $ ragon-admin audit egress ✅ Egress audit: pass Network destinations: 127.0.0.1:8000 29x model server 127.0.0.1:8767 1x Gmail search service DNS queries: none Networked git commands: none ``` Everything stayed on `127.0.0.1`. Not one DNS lookup. Need a hard guarantee? Type `/airgapped on` and every command the agent runs gets an empty network namespace from the Linux kernel. It isn't a polite request to the model: there is simply no network to reach. [Privacy & security →](docs/privacy.md) ### ⚡ Fast, with no rate limits and no bill A small drafter proposes 12 tokens at a time and the model checks them in one pass, with a 262K-token context. Run four agents at once and they all keep going: no quota, no "please wait", nothing to pay per token, ever. ### ✨ Nothing to configure - **No tuning.** The model, its context length, memory share, kernels and speculative decoding are chosen for the GB10 already. - **It won't freeze your machine.** On a GB10 the GPU and the system share one pool of memory, and an oversized model load can lock up the whole box. Ragon checks the host first and watches memory pressure during the load, stopping it before the machine stalls. - **Your other machines find it.** Run `ragon-admin node enable` on the GB10, install the same script on a Linux laptop or a Mac on your network, and `ragon` there finds the GB10 by itself. - **Updates are one command:** `ragon update`. ## What you get | | | |---|---| | 🖥️ **`ragon`, the terminal agent** | Reads your repository, runs commands and tests, edits code, and asks before anything risky. [More →](docs/ragon.md) | | 🌙 **Night Shift** | `/night add ` before bed; each task done on its own git branch by morning. Nothing is merged or pushed without you. | | 🔎 **Web search and fetch** | Through a private SearXNG on your own machine: no search account, no API key. | | 🧭 **A code index** | Definitions, callers and impact across your repository, so the agent finds code instead of grepping for it. | | 🧩 **Your existing skills** | Skills you already wrote for Claude Code, Gemini CLI, OpenClaw or Hermes work in Ragon unchanged. | | 💬 **Ask, in a browser** | `ragon web`: questions with web search, image search and dictation, and your agent sessions, from this machine or a paired phone. [More →](docs/web.md) | | 🪟 **Desktop app** | Ask and Work in a window of its own, with approvals, diffs and undo for agent sessions. [More →](docs/desktop.md) | | 📬 **Gmail, Drive, Calendar** (preview) | Read-only, through a sign-in that stays on your machine. Switched off entirely at `/airgapped on`. | ## Runs on every GB10 NVIDIA DGX Spark · Acer Veriton GN100 · ASUS Ascent GX10 · Dell Pro Max with GB10 · Gigabyte AI TOP ATOM · HP ZGX Nano · Lenovo ThinkStation PGX · MSI EdgeXpert They share the chip, the 128 GB of unified memory and DGX OS 7, so one installer covers them all. Ragon is developed and tested daily on the ASUS Ascent GX10; the others are covered by tests. ## FAQ **Do I need a GB10?** Yes, to run the model. Your other computers use it as clients: Linux (Intel/AMD or Arm) and macOS (Apple silicon or Intel), set up by the same install command. A Windows client (Windows 11 on Arm, such as the RTX Spark laptops, and x86-64) is available as an unsigned preview, installed with the PowerShell line above, and needs Smart App Control off. **Does `/airgapped on` work on a Mac?** Yes, enforced by macOS's own sandbox, with one difference: there the level is set when `ragon` starts (`ragon airgapped default on`, then restart), not switched inside a running session. **Is it really free?** Yes. Ragon is open source under the AGPL, and the model's weights are free. Your only running cost is the electricity. **What do I need besides the machine?** The OS it ships with (DGX OS 7, or Ubuntu 24.04 where the vendor offers it), Docker with the NVIDIA Container Toolkit, Python 3 and Git. On plain Ubuntu, `ragon-admin host check` tells you what's missing. **Does anything ever reach the internet?** Only what you or the agent asks for, such as a web search. The full list is below, and `/airgapped on` turns all of it off. **I found a bug, or something the egress audit flagged.** Please [open an issue](https://github.com/dreamference/ragon/issues). Reports from the audit are the most valuable ones we get.
What does reach the network, and when Local isn't the same as air-gapped, and Ragon doesn't pretend otherwise. These are the only things that leave the machine, and each one happens because you or the agent asked for it: | When | What is sent | To | |---|---|---| | Installing, first model start | Container images, packages, model weights | Docker registries, PyPI, Hugging Face, GitHub | | The agent or chat searches the web | The search query | Search engines, through SearXNG on your machine | | The agent fetches a page | A request for that URL | That website | | You connect Gmail, Drive or Calendar | Read-only requests | Google | | You run `ragon update` | A release check and download | GitHub | A search query is written by the model and can contain fragments of your context. At `/airgapped on` none of these happen: no search, no fetch, no mail, only the model on your machine.
The model Ragon serves one model, chosen and tuned for the GB10: **Qwen3.8-27B** in NVIDIA's 4-bit NVFP4 format (`qwen3.8-27b-nvfp4-dflash2`), served by ragon-serve (SGLang with `--engine sglang`) with the DFlash2 drafter, a 262K-token context and about 20 GB of weights. While it serves, roughly 38 GB of the machine's memory stays free. [More about the model →](docs/models.md)
Configuration Every setting resolves the same way: command-line flag, then `DREAMFERENCE_*` environment variable, then `dreamference.toml` (in the project, then `~/.config/dreamference/config.toml`), then the built-in default. ```toml # dreamference.toml vllm_host = "http://localhost:8000" model = "qwen3.8-27b-nvfp4-dflash2" ragon_airgapped = "on" # every session starts air-gapped ragon_gmail = false # never offer Gmail to the agent ``` The same model server also drives Cline, Continue and OpenHands: `ragon-admin run --agent cline "add type hints to utils.py"`. The web UI and the desktop app are set up in [Get started](docs/getting-started.md).
Build from source ```bash git clone --recurse-submodules https://github.com/dreamference/ragon.git ragon cd ragon python3 -m venv .venv && .venv/bin/pip install -e . export PATH="$PWD/.venv/bin:$PATH" # the agent runs ragon-admin, so keep it on PATH ragon-admin host setup # swap, kernel settings, out-of-memory guard ragon-admin server start # checks the host, downloads and loads the model ragon-admin codex build # compiles ragon and links it into ~/.local/bin cd ~/my-project && ragon ``` `ragon-admin` is the Python package that runs the model server, the sidecars and the builds ([command reference](docs/admin.md)); `ragon` and its web and code-index tools are Rust. Tests need no GPU or Docker: `.venv/bin/python -m pytest tests/`. Design specs are in [`specs/`](specs/README.md), and [How it works](docs/architecture.md) explains the pieces.
## Star history Star history of dreamference/ragon ## Spread the word Know someone with a DGX Spark? Send them this page. If you believe code should stay on the desk it was written on, **[star the repo](https://github.com/dreamference/ragon/stargazers)**: it's how other GB10 owners find Ragon. [Share on X](https://x.com/intent/post?text=Private%20AI%20coding%20agent%20that%20runs%20entirely%20on%20my%20DGX%20Spark.%20No%20cloud%2C%20no%20account%2C%20no%20telemetry.&url=https%3A%2F%2Fgithub.com%2Fdreamference%2Fragon) · [Share on LinkedIn](https://www.linkedin.com/sharing/share-offsite/?url=https%3A%2F%2Fgithub.com%2Fdreamference%2Fragon) Issues and pull requests are welcome, especially anything the egress audit turns up. ## Credits Ragon began as a fork of the open-source Codex CLI (Apache 2.0) and is growing into the world's leading confidential local AI software. It also stands on [SGLang](https://github.com/sgl-project/sglang), [vLLM](https://github.com/vllm-project/vllm), [SearXNG](https://github.com/searxng/searxng) and the [Qwen](https://github.com/QwenLM) models. Thank you to all of them. ## License Copyright (C) 2026 Dreamference contributors. Ragon is free software under the **GNU Affero General Public License v3** or later; the full text is in [`LICENSE`](LICENSE). Section 13 is the clause that matters for a fork: if you run a modified Ragon and let people use it **over a network**, you must offer them its source. This covers Ragon's own code. The components it deploys keep their own licences: Codex (Apache 2.0), SGLang and vLLM (Apache 2.0), and the model under its own weights licence. NVIDIA, GB10, DGX and DGX Spark are trademarks of NVIDIA Corporation. Qwen is a trademark of Alibaba Cloud. Other names are trademarks of their owners. Ragon and Dreamference are not affiliated with, sponsored by or endorsed by any of them; the names say only what Ragon runs on and is built from.