Features
Everything you need. All in one place.
From chat-to-code to one-click deploy — fully integrated, no tool switching.
01
AI & Code Generation
You watch the code appear — line by line, instead of a wall of text after 30 seconds.
The agent builds iteratively — spotting errors itself and fixing them without you asking.
Drop in a screenshot — the agent reconstructs layout and styling faithfully.
When the build breaks, the agent explains why and proposes a fix.
Up to 1,000 autonomous iterations per task — testing, debugging, improving. Stop the agent anytime.
TS errors, missing imports, failing tests — fixed automatically.
The newest model generation in a redesigned picker — search, filters, plain language.
Queue your next prompts while the agent works — pause, reorder or remove them anytime.
Conventions and dos & don’ts the AI follows in every single turn.
The assistant sketches self-contained UI prototypes in their own tab — “Build this” turns one into a real change.
Hand over a URL and the assistant captures the site — pages and assets land in your project, with live progress in the chat.
Documents, archives, data files — each with a preview card that recognises the type. Upload limits and safety checks stay in place.
02
Developer Experience
Read-only code view built on Monaco — the engine behind VS Code — with syntax highlighting.
Auto-commits, branches, pull requests — or auto-sync with your own repo.
Changes appear in under 200ms. Zoom, simulate devices, toggle design mode.
Click any element — change color, text, Tailwind class. The change lands in the code.
Console, network panel, backend logs and Swagger UI — right in the browser.
Full version history. One step back, three steps back, or a complete revert.
A Tests tab where the AI writes and runs tests — results per case, coverage and history.
Publish, build and task alerts via email, Slack, Discord or Telegram — per event and channel.
A clean copy — code, database, integrations and a fresh git repo, live status included.
A docs/ folder next to frontend/ and backend/, shown in its own tab: markdown rendered, source one toggle away, versioned with the code.
03
Data & Backend
Every project gets its own DB. No setup, no manual migrations.
A real FastAPI backend in Python — readable, exportable code, not a no-code black box.
Query, edit, export — all without pgAdmin.
S3-compatible storage with automatic CDN URLs.
Secrets per environment (Dev / Prod), encrypted at rest.
04
Publishing & Deployment
Click → live on *.airbuild.space with CloudFront CDN and TLS.
Your own domain with automatic SSL (AWS Certificate Manager). Pro plan and up.
Optional dedicated container per app with its own CPU, RAM and disk limits (Power) — six tiers from XS to 2XL.
Every publish attempt is recorded with its commit and gate results, so you can see exactly what shipped and when.
Search the live logs of every published app and jump from an error straight into a fix.
Email reminders before a hosted app is taken offline and before its data is deleted.
Every publish attempt with its own version number, full check log and findings in Problems.
Code quality, dependency, security and API-lint checks run before every publish.
Shared cloud from €5/mo or switch to dedicated private hosting with fixed CPU, RAM and database storage in its own container.
Every project owns a target from day one — inactive and free — and name, hosting and SEO title are required before it can publish.
Pick which pre-release findings to address, fix them with AI, then resume the held publish.
05
Auth & Security
Email and password generated straight into your backend — Argon2id hashing, JWT sessions, protected routes. Google and GitHub sign-in via a Supabase project you connect.
TOTP two-factor authentication for your Airbuild account, with single-use backup codes — for email/password logins; Google and GitHub accounts are secured by that provider.
OIDC and SAML 2.0 per organisation — any standards-compliant IdP, e.g. Entra ID, Okta, Google Workspace. Suite Builder plan.
Org-wide AGENTS.md rules plus control over which AI models members may use.
Clear AI labeling plus machine-readable marking in published sites and generated images.
Flags API keys, tokens and connection strings in chat messages and offers to store them as an environment variable — Trivy re-scans the source for hardcoded secrets before every publish.
You decide which third-party scripts and API origins your published app may load — enforced at publish.
Accounts are created just in time on the first SSO login, restricted to email domains you have verified with us — Suite Builder plan. SCIM 2.0 is implemented but not yet exposed.
A monthly AI-task budget per member, tracked against their own limit — one member cannot consume the whole shared pool.
FAQ
Frequently asked.
Which AI models can I build with?
Airbuild supports leading models from Claude, OpenAI, Google and Mistral — including Claude Fable 5, Opus 5 and Sonnet 5, OpenAI GPT-6 Astra and the GPT-5.6 variants, and Google Gemini 3.7 Flash and 3.1 Pro — plus open-weight models (DeepSeek, Kimi K3, Qwen3-Coder, GLM-5.3), EU-hosted by default and available natively from their own provider if you prefer. You choose the model per project, and on team plans Org Rules let you control which models are allowed for your projects.
What backend and database do I get out of the box?
Out of the box, a Web App project comes with an auto-provisioned PostgreSQL database and a full Python/FastAPI backend with email-and-password authentication, plus S3-backed file storage and encrypted per-environment secrets, with no separate setup. The backend is one coherent application rather than a set of short-lived functions, and every deployment target gets a database of its own.
Can Airbuild detect and fix its own errors?
Yes — Airbuild detects its own errors and usually fixes them before you see them. The Build-Failure Analyzer and self-healing catch build and startup errors during generation and repair them automatically, while the build is still running. Runtime errors in the live preview get a one-click AI fix. The result is far fewer broken deploys, and every repair stays in the project history.
How does Airbuild keep generated code secure and production-ready?
Airbuild keeps generated code secure and production-ready with a gate that runs before every publish: automated security and quality scans with one-click AI fixes, plus secret detection and transparent release checks. High-severity findings block the publish. The generated test suite can be switched on per project as an additional gate, and teams get SSO/SAML and EU AI Act support on top.