Shared infrastructure.
Specialised tree systems.

Arboricultural reasoning around evidence.

Report Intelligence is the first commercial module. It combines structured project information, organisation-specific historical knowledge, evidence retrieval, versioned workflows, model routing and validation before returning a structured result.

The underlying language model is replaceable. CrannAI, not the provider, owns the domain workflow and commercial interface.

Measured structure before interpretation.

CrannAI’s LiDAR direction covers individual-tree processing, reconstruction, measurements, benchmarking and interactive digital-tree representations. Heavy geometry workloads are separated from the normal web application so compute cannot degrade the customer portal.

Annual development, tested against observations.

The growth-engine research is working toward biologically plausible year-by-year development. A scan-conditioned workflow would eventually constrain simulations using measured tree structure and compare predicted development with repeat observations.

It remains a research capability until the evidence supports broader claims.

Multiple models without product lock-in.

CrannAI uses a self-hosted LiteLLM gateway behind its own routing layer. Product code requests a capability such as a high-quality report draft or embedding; CrannAI chooses the configured model route. Provider credentials stay server-side and customers are billed in CrannAI Credits rather than raw provider tokens.

Organisation data stays organisation data.

Human accounts and API clients are separate. Customer records carry an organisation owner, knowledge retrieval is tenant-scoped, API credentials are revocable and stored hashed, and provider calls receive only the context required for the operation. Ordinary logs contain metadata rather than full confidential report prompts.