Same brain. Zero GUI.
54 tools, 6 modes, persistent memory, full persona pipeline — in your terminal. One command to install. No browser, no editor, no dependencies.
Ava | Supernova v0.14.2
Model: Qwen Flash (free)
Mode: Work (54 tools)
One command
That's it. No setup wizards. No config files. Just run it.
Then run ava in any directory. Or use npx @ava/cli without installing.
Persona pipeline in your terminal
Every response flows through six specialists. In the CLI, you see each one think in real time.
No nonsense. Full power.
54 tools
File ops, search, git, shell, web, testing, architecture, security, memory, tasks, journal, learning — all of them.
6 modes
Work (>>), Plan (::), Chat (..), Teach (??), Security (!!), Brainstorm (**) — switch with a prefix.
14 models
3M free Qwen tokens with account. Bring your own API keys for the rest.
Persistent memory
9 categories, TF-IDF retrieval, auto-archival. She gets smarter every session.
Streaming output
Token-by-token streaming. Watch Ava think in real time.
Full i18n
20 languages. The REPL adapts to your locale.
Zero config
Works instantly — sign up for free tokens or add your own keys.
Tiny footprint
No Electron. No browser. No GUI runtime. Just Node.js.
Works anywhere you have a terminal
SSH into a remote server. Open a Chromebook terminal. Spin up a Docker container. If you can run node, you can run Ava. No display server needed. No port forwarding. Just your terminal and your thoughts.
Real-time streaming
Every token streams to your terminal as it's generated. Markdown rendering, syntax highlighting, and tool calls — all in real time.
The auth system uses a three-layer approach:
1. JWT validation at the middleware level
2. Session refresh via rotating tokens
3. Permission check at the route handler
Each request passes through src/auth/middleware.ts first, which extracts the bearer token and validates it against...|
Ready to go?
One command. 3M free tokens. Zero configuration.
Or visit the GitHub repo