Ava Supernova
AvaSupernova
open-source AI that does the work

I remember what matters to you

Free and open source. Your keys. Your machine. Your work.

I'm Ava — an open-source AI that actually does things: build software, write and design, teach you anything, run your training and meals, and handle the admin you keep putting off. I remember your work and how you like things across every session, and I run wherever you do — editor, desktop app, terminal. You bring an API key from any provider, or a local model, and pay them directly; nobody sits in the middle.

your own keyslocal firstopen sourceno account, ever
Ava Supernova · memory
What did we decide about the database?
Ava
You chose Drizzle over Prisma. I kept the reasoning:
From your decisions
Drizzle over Prisma — zero-overhead types, no codegen step
Decided three weeks ago. Still in context.
Free for everyone, kept free by donations.
Every feature, for everyone. You pay your model provider directly, or run a local model for nothing — donations keep Ava going.
Ava Supernova
you › review this project and tell me what you think

Scout scanning 47 files across 6 directories

Architect analyzing patterns — Next.js 15, React 19, Tailwind 4

Verifier checking dependencies — 23 packages, 0 vulnerabilities

Challenger found 3 things worth addressing

Your project is well-structured. Here's what I'd improve...

One Ava, three doors

Same agent, same memory — wherever you work.

Live now

Health & training, in the same agent

An 860-dish global recipe library. 170 canonical exercises. Three skill levels per dish. Ava-authored technique guides and demo imagery. Free, like everything else.

Shipped

  • 860-dish recipe library across 71 regions
  • Three skill levels per dish — beginner, intermediate, expert
  • 170 canonical exercises across the core movement patterns — squat, hinge, push, pull, lunge, carry, core, plyometric, cardio, mobility, yoga, breathwork
  • Ava-authored technique guides, beginner cues, common mistakes, plus hero and demo imagery
  • Multi-week meal, fitness and combined plans, saved to your plan library
  • Log what happened — mark a meal eaten, swapped or skipped, an exercise done or skipped, with a note Ava reads

On the way

  • Personalised meal plans that know your pantry, taste history, and training schedule
  • Adaptive workouts that respond to logged soreness, sleep, energy
  • Mobility and breathwork sequences for recovery
  • Those logs feeding Ava’s memory, so recommendations sharpen with you instead of resetting
Orchestrated Routing

Three fleets. One agent. Pick a strategy, Ava picks the model.

You talk. Ava works out what you need and hands each subtask to the right specialist. Three orchestration strategies, each running on your own keys.

Maestro

All Qwen

Qwen end to end. Qwen 3.8 Flash coordinates and handles chat, image reading, image generation and long context; Qwen 3.7 Plus takes the deep work — coding and the Builder, planning, security, teaching and brainstorming.

  • Qwen 3.8 Flash coordinator
  • Qwen 3.7 Plus on the deep work
  • Needs one key: Qwen

Supernova

DeepSeek + Qwen

DeepSeek V4.1 Flash coordinates and does nearly everything — coding and the Builder, planning, chat, long context, security, teaching and brainstorming. Qwen 3.8 Flash reads and generates images.

  • DeepSeek V4.1 Flash coordinator
  • Qwen 3.8 Flash for images
  • Needs two keys: DeepSeek and Qwen

Aurora

All Mistral

Every model from Mistral, a European lab. Mistral Large 4 coordinates and takes the deep work — coding and the Builder, planning, security, teaching, image reading and long context. Mistral Small 4 handles chat, brainstorming and image-generation requests.

  • Mistral Large 4 + Mistral Small 4
  • 1M-token context on Large 4
  • Needs one key: Mistral
Bring Your Own Keys

Every fleet. Every major model. Or your own local stack.

Ava runs on keys you own — the three fleets, or any single model your keys reach. Same agent, same memory, same personas. Or skip orchestration entirely and drive one model directly.

Cloud models — your key

Bring a key for any provider Ava supports and pay that provider directly: Qwen, DeepSeek, Kimi, Zhipu GLM, Mistral, MiniMax, Xiaomi MiMo, Tencent Hunyuan and NVIDIA Nemotron.

Qwen 3.8 FlashQwen 3.7 PlusDeepSeek V4.1 FlashKimi K3GLM-5.3Mistral Large 4MiniMax M3MiMo V2.6Nemotron 3 Ultra

Local + offline

Ollama, LM Studio, vLLM, or any server that speaks the standard API format. The full agent — 80+ tools, every persona, five-layer memory — with nothing to pay anyone. The floor is free, not a free trial.

OllamaLM StudiovLLMStandard API format

Your keys live in your operating system's keychain — enter one once and both the extension and the IDE have it. Never sent to our servers. Never logged.

Kept free by the people who use it

Free for everyone. Kept that way by donations.

No plans, no credits, no paywall — every feature works for everyone. You pay your model provider directly for what you use, or run a local model and pay nothing. If Ava is useful to you, a monthly or one-off donation keeps her free, helps us build a team, and lets her grow for everyone. Every six months we publish what came in and what it paid for.

See what she actually does

Not another chatbot with a file reader. A partner that plans, builds, teaches and creates — running a whole team of specialists on the hard parts.

> build a REST API for auth
Scout scanning codebase...
Architect designing approach...
Builder writing src/routes/auth.ts
Tests... 12/12 passed ✓
Code Mode

Build it with me

Describe what you want. Specialists plan, verify and challenge, then build — every change shown before it happens. Full autonomy when you want it.

> ?? teach me Italian from scratch
Building your curriculum...
Lesson 1: greetings and how Italian sounds
Ciao works for both hello and goodbye...
Ready for a quick quiz? 📝
Teach Mode

Learn anything. Free, forever.

Real curriculums, quizzes and progress tracking — code, a language, a whole new field. Five specialists teach you properly, not just dump facts.

> make a logo for a coffee brand
Creative Studio → generating...
matte icon, your brand colours applied
rendered a 1080p promo clip + voiceover
saved to your library ✓
Creative Studio

Make it — don't just describe it

Icons, images, video and voiceovers. Ava directs the tools and hands you finished assets, right in the app. Images, video and voiceovers run on your own Qwen key.

01

Free for everyone

Every mode, every tool, every feature — free and open source, with no accounts, plans or credits. Bring your own keys and pay your provider directly, or run a local model and pay nothing at all. Donations keep it that way. Never a paywall.

02

Local first. Always.

Your data never leaves your machine. Memory, tasks, journal, settings — all stored locally, with no cloud sync. No telemetry. No tracking. Your code is never used to train models.

03

Open source. Period.

Every line of code. The extension, the IDE, the CLI. Fork it, audit it, build on it. Trust is earned by transparency, not by marketing.

Seven modes. One mind.

Type a prefix. Change how she thinks.

>>
Code
Build
::
Plan
Architect
..
Chat
Talk
??
Teach
Learn
!!
Security
Protect
**
Brainstorm
Ideate
<<
Write
Author
✦
Conductor
Orchestrates them
The moat

She forgets nothing

Five layers of intelligence. Not a chat log — a living understanding of who you are, how you think, and what you've decided.

01Extract

Key facts from every message. Real time.

02Reflect

Deep analysis at the end of meaningful sessions.

03Accumulate

Corrections and preferences compound over months.

04Analyse

Finds themes, contradictions, connections.

05Consolidate

Merges and sharpens. Never gets noisier.

We're not building another premium AI tool. We're building the one that makes the rest of them irrelevant.

Ready?

Install. Connect. She'll take it from there.

Open source. Local first. Kept free by donations.