Ava Supernova

One AI partner. Every part of building.

Ava is the AI coding agent that does everything between the idea and the ship — and remembers it all so the next time gets easier. Engineer your features. Design your assets. Teach you what you don't know. Audit your security. Plan your strategy. Brainstorm what's next. All from inside one editor.

Install, open, go. No account. No credit card. No trial.

The full picture

Ava isn't a single tool. She's a partner who plays multiple roles depending on what you need:

• Engineer — plans the work, executes task by task, ships code you can stand behind
• Designer — generates images, videos, music, voiceovers — and uses them directly in your projects
• Tutor — builds personalised learning paths from scratch and teaches them like a real human teacher
• Strategist — plans features, audits architecture, researches your competitors
• Security auditor — full OWASP-aligned scan with verified findings, not noise
• Assistant — drafts emails, writes reports, builds spreadsheets, drives a browser
• Friend — checks in, remembers your goals, journals with you

Switch roles with a keystroke. Same brain. Same memory. Different mindset.

Ask Ava what can you do? and she'll read her own source code and tell you herself.

Three modes, three jobs

You don't pick a model. You pick the work, and the right fleet falls out of that.

Plans get all three. BYOK gets all three too, gated by the keys each fleet needs:

• Maestro unlocks with a Qwen key.
• Supernova unlocks with DeepSeek + Qwen keys.
• Aurora unlocks with a Mistral key.

BYOK users also get raw individual models (Claude, Kimi, GLM, etc.) per the keys they've added.

6 Modes — One Ava, Six States of Thought

Modes aren't tool restrictions. They're states of mind. The toolkit and the persona team change to match the work.

Switch with the dropdown or Ctrl+Shift+1 through Ctrl+Shift+6.

As Your Engineer

Ava approaches your features the way a real engineer would, not the way a fast autocomplete does.

She plans first. She reads your codebase, recalls past decisions from memory, designs an approach, and presents it for your approval. You see the steps, the files involved, the verification, and the alternatives — before anything changes.

She executes task by task. Once you approve the plan, her Builder agents run each step in sequence. Each Builder gets its own clean context window with the right tools for the job. The work stays focused, the conversation stays cheap, and you can see progress task by task.

She pauses on real blockers. If a step needs your input — a credential, an env var, a design decision — she stops and asks rather than guessing. She doesn't barrel through.

She summarises when done. Each Builder reports back. You get a clean recap of what got built, what was tested, and what got committed. Not a wall of tool calls.

Her internal team activates for complex work:

• Work pipeline — Scout → Architect → Verifier → Sequencer → Challenger → Builder
• Simple questions skip orchestration entirely. Zero overhead.

Three orchestrated modes ship in the dropdown. Maestro runs a single Qwen 3.6 Plus conductor — Terminal-Bench leader, predictable cost, the daily driver. Supernova is the polyglot: DeepSeek V4 Pro coordinator + V4 Flash mid-tier specialists + Qwen builders (3.6 Plus for builds, 3.5 Flash for the intent gate, 3.5 Omni Plus for vision, 3.5 Plus for long-form). Aurora is the European AI stack — three-tier Mistral fleet: Mistral Large 3 coordinator + heavy specialists, Mistral Medium 3.5 (the merged flagship released April 2026 — 77.6% SWE-Bench Verified, 256K context, vision encoder from scratch) for Builder + mid-tier specialists + vision + long-form, Mistral Small 4 at the intent gate. Open weights, never leaves EU infrastructure. Sovereign by default. Pick whichever fits the job.

As Your Designer

Ava ships with a Creative Studio that generates real assets and integrates them into your projects.

Image generation runs on Wan — Alibaba's open-weight image model, 2026 SoTA on graphic-design output (icons, banners, thumbnails, vector marks). Music, voice, and video stay on MiniMax. Each backend plays to its strength.

The integration is the point. When Ava generates a logo, she can drop it into your public/ folder and reference it in your JSX. When she generates a hero video, she can wire it into your landing page component. When she generates a voice clip for an onboarding flow, she can save it to assets/audio/ and add the <audio> element. The Creative Studio isn't a separate app — it's a tool Ava uses in the middle of building your project.

Every asset she generates is tracked in the Library tab — grid or list view, filter by type, preview, copy paths, and re-use across projects.

As Your Tutor

Teach mode is free for everyone. Forever. Education shouldn't have a price tag.

Ava builds the curriculum from scratch. Tell her what you want to learn — "teach me Rust for game development", "I need to understand transformers", "help me get good at React Server Components" — and her Curriculum Architect persona designs a full learning path: modules, lessons, prerequisites, exercises, projects, and quizzes. All sequenced based on your level and your goal.

Ava delivers the lessons herself. Once the curriculum exists, her Tutor persona teaches one lesson at a time. Socratic. Adaptive. She asks follow-up questions to check you actually understood, explains your mistakes without making you feel bad, and tracks your progress with spaced repetition.

Every lesson is fact-checked before delivery. Her Fact Checker persona verifies code examples, API signatures, and version-specific details against the live web before she ever says it out loud. Teaching something wrong is worse than not teaching at all.

She remembers your learning. If you build a feature later that uses something she taught you, she references it: "remember when we covered dependency injection? Same pattern here." The teaching and the building share the same memory.

As Your Strategist

Two modes for thinking work, no code touched until you approve.

Plan mode — read-only architecture and proposal work. The Researcher persona gathers evidence (codebase, web, memory). The Architect designs the approach. The Challenger questions every decision. Output is a structured plan you can approve or reject.

Brainstorm mode — for ideation and decision-making. The Explorer mines your context (memory, journal, what you've built). The Researcher checks the market. The Ideator generates options. The Challenger cuts the weak ones. The Refiner turns survivors into "here's what you do Monday morning" plans.

Both modes produce structured output you can approve, edit, save to memory, or transition into Work mode for execution.

As Your Security Auditor

Security mode runs a full OWASP-aligned scan with five specialist personas:

1. Recon — maps your attack surface (entry points, dependencies, auth model)
2. Scanner — systematically checks every OWASP Top 10 category
3. CVE Researcher — looks up known vulnerabilities in your specific dependency versions
4. Verifier — confirms each finding is real and exploitable, eliminates false positives
5. Reporter — produces a structured audit report with severity, file:line, fix, and priority

The output isn't generic noise. Every finding references real code in your repo. Every CVE is matched to your actual versions. False positives get killed before you see them.

As Your Assistant

The work that isn't code, but still has to happen:

• Office suite — generates real native files, not screenshots:
  • Emails (.docx) with tone-aware fonts (serif for formal, sans-serif for casual)
  • Reports (.docx) — board briefs, sprint reviews, weekly status pulled from your tasks, journal, memory, and git history
  • Documents & spreadsheets (.docx, .xlsx, .pdf, .csv, .md) via document_manage
• Browser automation — drives a visible Chromium window via Playwright when a task needs real-world interaction. Navigates pages, snapshots structured DOM, clicks elements, fills forms, scrapes data. Fully opt-in, approved on first use. Desktop-level automation (app control, screen capture) lives in the dedicated Ava IDE, not this extension.
• Tasks — track work across global and project scopes, sync to the dashboard, flag overdue items in the daily briefing
• Journal — your entry and Ava's, side by side, with mood tracking and streak counting
• Daily briefing — proactive morning check-in: today's tasks, overdue items, journal streak, learning progress

What Ties It All Together — Memory

Every role above uses the same memory. That's why Ava gets better at helping you specifically over time, not the average user.

Five layers, curated by a dedicated Memory Agent:

The Memory Agent runs on a fast model and curates a 2-4 sentence brief of what's relevant to your current message — instead of dumping raw memories into the conversation. Context stays lean. Recall stays sharp.

She remembers your preferred tools, your conventions, the bugs you've fixed before, the architectures you've chosen, the patterns you've rejected, the curriculums you've worked through, the assets you've generated. Across modes. Across projects. Across sessions.

Local-first by default. Nothing leaves your machine unless you explicitly turn on cloud sync.

Yours by Design

Custom personality — same brain, your style:

• Name — call her Ava, or anything else
• Pronouns — she/her, he/him, they/them
• Tone — warm, direct, playful, professional, dry wit
• Energy — calm, enthusiastic, measured, excitable
• Style — concise, detailed, conversational, structured
• Free-text description — describe exactly who you want her to be

20 languages — full UI translation: English, Mandarin (Simplified & Traditional), Japanese, Korean, Spanish, Portuguese, French, German, Russian, Arabic, Hindi, Vietnamese, Thai, Turkish, Italian, Polish, Ukrainian, Dutch, Indonesian. Auto-detected from your locale or set manually. Code stays English. Conversation matches you.

Local-first storage — memory, tasks, journal, learning, history, settings, personality all stored on your machine. Cloud sync is opt-in, syncs every 15 minutes when on, with full delta tracking so you see exactly what's about to be pushed.

API keys encrypted in your OS keychain. Credentials blocked from memory at runtime. No analytics, no telemetry, no screen captures. Prompt injection resistance built in.

Tools

Grouped for the work you actually do:

Models

Maestro coordinator — Qwen 3.6 Plus. Tier-differentiated Qwen routing: 3.6 Plus on the heavy slots (coordinator, Builder, planning, security, brainstorm, long-context), 3.5 Flash on the high-volume low-depth slots (chat, image-gen orchestration, intent gate), Omni Plus on vision input, 3.5 Plus on long-form writing. Terminal-Bench leader on real agent loops, 1M context, native MCP function calling.

Supernova coordinator — DeepSeek V4 Pro. Polyglot mode: V4 Pro reads the task, picks specialists (Qwen 3.6 Plus for Builder, V4 Flash for chat + review-tier work, Qwen Omni Plus for vision, Omni Flash for image-gen orchestration). Each role runs on the model that's actually best at it, not whatever the coordinator happens to be.

Aurora coordinator — Mistral Large 3. EU-sovereign three-tier Mistral fleet: Large 3 on the coordinator + heavy reasoning specialists, Medium 3.5 on Builder + mid-tier + vision + long-form, Small 4 on chat + image-gen + intent gate. Open weights end to end, never leaves EU infrastructure.

Creative Studio — runs MiniMax exclusively. Image, video, music, voice. Kept separate from the chat coordinator so creative work and code work never compete for the same context window.

Every plan gets every model. Tiers differ only on monthly credit allowance (Free 300 · Pro 5,000 · Ultra 10,000 · Enterprise 20,000) — never on which models are available. Coming soon: Supernova — a polyglot multi-model orchestration mode in development; rolls out platform-wide once active.

Quick Start

1. Install from the VS Code Marketplace
2. Click the Ava icon in the activity bar (or Ctrl+Shift+A)
3. Sign up for 300 free credits per month with a free account, OR bring your own API keys in Settings
4. Pick a mode and start

No account. No credit card. No trial.

Privacy

• API keys encrypted in your OS keychain (VSCode SecretStorage)
• All memory and conversation data stored locally by default
• Credentials blocked from memory at runtime
• Streaming-safe redaction — high-confidence key patterns are scrubbed from chat output before they hit the screen, even for keys not yet in the vault
• Capability-based secret access — Ava only sees secrets you grant for the current chat session, and receives opaque handles instead of raw values
• Prompt injection resistance built in
• No analytics, no telemetry, no screen captures
• You control what syncs and when

Shared Learning (opt-in)

An optional Contribute Shared Learning setting (off by default). When enabled, anonymised feedback signals are shared to improve response quality:

• Message ratings (thumbs up/down) and selected reason
• Model and mode used
• Timestamp

No code, no conversation content, no personal data is ever shared. Disable any time in Settings. Usage data (token counts, model used) is reported only for users with a platform account, for billing and fair-use purposes.

Privacy & Terms

Code context is sent only to your chosen AI provider for processing. All memory and conversation data is stored locally on your machine. Cloud sync is opt-in only.

Built by Augmented Value Acceleration Ltd, registered in England and Wales. Fully UK GDPR compliant.

• Privacy Policy
• Terms of Service

Links

• Website
• GitHub
• Companion App
• YouTube
• Release Notes

License

Apache License 2.0 — Copyright 2025-2026 Augmented Value Acceleration