How FrootAI Works
/engineTwo engines, one pipeline. The Harvest engine turns any GitHub repo into a deployable Solution Play. The AVM Intelligence enginecomposes verified Azure infrastructure from Microsoft's AVM module catalog. Both are audit-ready by default.
0/0
0/0
0
—
—
Harvest Engine — 7-Stage Pipeline
From GitHub URL to deployable Solution Play in 7 stages.
Discover
Turn any GitHub repo URL into a validated discover-record with full metadata, license, and provenance — zero LLM calls.
Fetch
Clone, filter, and prepare the repo content for extraction — language detection, file-tree pruning, binary exclusion.
Extract
LLM-powered fact extraction — architecture, services, regions, models, deployment patterns into structured RepoFacts.
Retrieve
Semantic retrieval against the 101-play reference corpus — find the most similar existing plays for scaffolding context.
Scaffold
Generate the Solution Play scaffold — agent.md, architecture.md, README.md, deploy scripts — all fact-checked against RepoFacts.
Compose Infra
Compose verified Azure infrastructure from AVM modules — Bicep templates matched to the play's resource requirements.
Customize
Apply policy overlays (naming, tagging, networking, RBAC) to the composed infrastructure via a YAML-driven customization pass.
AVM Intelligence — 5-Component Chain
Verified Azure infrastructure composed from Microsoft's AVM module catalog. Five components, each purpose-built for a stage of the composition pipeline.
AVM Catalog
Live catalog of every Azure Verified Module — Bicep + Terraform. Version-pinned, weekly-refreshed, popularity-ranked.
WAF + CAF Pack
80+ WAF checks + 60+ CAF checks as executable validators. Per-pillar scoring drives marketplace badges.
Resolver
Intent-to-module resolution. Given a resource intent, return ranked AVM module candidates with reason + selection.
Composer
Compose verified Bicep infrastructure from resolved AVM modules — dependency-aware, parameter-wired, ready-to-deploy.
Validator + Emitter
Validate composed infrastructure against WAF/CAF rules, compute cost + carbon estimates, emit the final verified template.
Why This Wins — 5 Asymmetries
Structural advantages the harvest + AVM intelligence engines create that competitors cannot replicate without comparable investment.
We own the provenance trail nobody else publishes
Cursor, Lovable, and Bolt ship generated IaC without telling the user which AVM module backs each resource. FrootAI exposes composed_from, confidence, upstream commit SHA, and cost per play inline. A CIO clicking "How was this made?" gets the audit-ready answer no competitor can give.
Browse plays with provenance →Shipping cadence as marketing
Public release notes, a live roadmap, and an auto-generated changelog turn engineering shipping cadence into product trust — zero marketing content needed.
See the changelog →Studio is the only try-before-buy in the category
Competitors require sign-in + credit card before showing the product. FrootAI Studio lets any visitor browse the AVM catalog, try a composition, and see real cost projections in 60 seconds — no sign-in for read-only flows. The trajectory projects +25% paid-tier sign-ups attributable to Studio touch.
Try Studio →Lab benchmarks become reference material
Posts like "AVM-composed Bicep vs hand-authored: cost across 100 workloads" become citations in blog posts, conference talks, and academic papers. Procurement teams and Microsoft FastTrack partners forward Lab posts. That's earned distribution competitors cannot replicate without comparable engine investment.
Explore the Lab →Schema-on-disk doubles as machine-readable API
The 7 harvest schemas + AVM module-record schema + WAF/CAF/cost/carbon report schemas live at /schemas. Partners validate their own pipelines against our schemas — distribution multiplier into FastTrack, SI partners, and cloud-cost SaaS tools. Every schema is CC0-licensed for maximum adoption.
Browse schemas →