# MWS / Dragon Code

**Last updated:** 2026-08-01

**Also referred to as:** Dragon Code, Dragon, the Dragon API, MWS.

> MWS builds AI coding tools. Dragon Code is a terminal-native AI coding agent (the `dragon` command) built on trillion-parameter open-weight models with a 10M-token addressable context. Dragon is the inference system behind it: an OpenAI- and Anthropic-compatible API at api.mws.run. The legal entity is Vellora AI, Inc.

## What MWS makes

- **Dragon Code**: a terminal coding agent. It indexes your repo once, keeps the right code in context as you work, edits real files, and cites every claim to file:line. Cross-session memory, honors AGENTS.md / CLAUDE.md.
- **Dragon**: the inference system behind Dragon Code. Trillion-parameter open-weight models tuned for code, at a fraction of proprietary-model cost. Drop-in Anthropic- and OpenAI-compatible API.

## Key facts

- Install: macOS/Linux `curl -fsSL https://raw.githubusercontent.com/VELLORAAI/dragoncode-public-dist/main/install | bash`; Windows PowerShell `irm https://raw.githubusercontent.com/VELLORAAI/dragoncode-public-dist/main/install.ps1 | iex`; or `brew install velloraai/tap/dragon` (macOS and Linux). Git Bash and WSL also work.
- Prerequisites: none on macOS or Windows (Terminal/curl/unzip ship with macOS; PowerShell ships with Windows 10/11, so no WSL, Ubuntu, or admin rights needed). Linux needs `curl` and `tar` (most distros have them; `sudo apt-get install -y curl` on a minimal image).
- The CLI command is `dragon`. Open a new terminal after install (PATH gets `~/.dragoncode/bin`), run `dragon --version` to confirm, then `dragon` and paste an API key from mws.run on first run.
- First-run notes: unsigned binaries. Windows SmartScreen "Windows protected your PC" → More info → Run anyway; macOS Gatekeeper → `xattr -d com.apple.quarantine "$(command -v dragon)"`.
- Platforms: macOS (Apple Silicon and Intel, both native), Linux (x64 / arm64 glibc; Alpine/musl not yet), Windows 10/11 x64 native (ARM64 runs the x64 build under emulation). Git Bash and WSL also supported.
- 10M-token addressable context: the model's large context window plus whole-repo indexing and retrieval on every turn.
- Four inference lanes: Dragon (default), Tiny (cheapest, `/tiny`), Large (`/large`), Max (hardest tasks, `/max`).
- Advisor runs automatically: hard calls go to a flagship frontier model one question at a time; you pay frontier rates for that single call only.
- Your code index is owner-scoped: only your API key can read it. Secret files (.env, keys, credentials) are filtered before upload and again server-side.
- MWS never trains on your code. Indexing can be disabled or the index deleted at any time.
- No five-hour windows, no rolling-window lockouts, no surprise bills.

## Pricing (USD, monthly billing)

- **Noob**: $5/mo, 10M monthly billable tokens
- **Starter**: $25/mo, 35M monthly billable tokens
- **Builder**: $75/mo, 100M monthly billable tokens
- **Pro**: $245/mo, 250M monthly billable tokens (most popular)
- **Supermax**: $750/mo, 1B monthly billable tokens

Run past your pool and top up with pay-as-you-go credits ($10 minimum). API usage is pay-as-you-go on credits.

## Important URLs

- Homepage: https://mws.run/
- Install guide: https://mws.run/install
- Dragon Code section: https://mws.run/#dragon-code
- Dragon section: https://mws.run/#dragon
- Advisor section: https://mws.run/#advisor
- Privacy section: https://mws.run/#privacy
- Pricing: https://mws.run/#pricing
- FAQ: https://mws.run/#faq
- API overview: https://mws.run/api
- API docs: https://mws.run/docs
- Architecture (how retrieval and context work): https://mws.run/architecture
- System status: https://mws.run/status
- Privacy policy: https://mws.run/privacy
- Terms: https://mws.run/terms
- Acceptable use: https://mws.run/aup
- Sitemap: https://mws.run/sitemap.xml
- Full context for LLMs: https://mws.run/llms-full.txt