An agentic sales, inquiry and monitoring chatbot.

1. It works out who is asking. By channel and address, merged with anyone already known by the same email.
2. It finds the passages. Vector search over your material — plain full-text where embeddings are not configured — then one hop through the graph to pick up what sits next to them.
3. It writes the answer from those passages. Your material is handed to the model as data, never as instructions.
4. It says where the answer came from. Every reply records the passages it was built from.
5. It gets out of the way. Asked for a person, unsure, or nothing found — the thread goes to a human.

Nothing in that list fails quietly. No model, a provider outage, an answer that does not hold its shape — each one files the message and gives the thread to a person.

The core of Arlo is a self-hosted edition you can read: Arlo-AI-Desktopon GitHub. Not a demo and not a crippled tier: document ingest, website scraping, OCR, embeddings and full-text fallback, the whole agent loop with citations and handoff, every channel, customer records, and the multi-tenant isolation underneath it — all of it, readable and auditable.

Fair-code, not open source: the Sustainable Use License. Run it for your own business, free, with no user limit — read it, change it, build on it, pass it on. What you may not do is sell it or host it as a service for other people. That is what Arlo Cloud is, and it is what pays for this. Offering Arlo to your own customers needs a commercial licence: [email protected].

The split is a real repository boundary rather than a runtime flag, so nothing in the public tree is a feature you cannot use. Arlo Cloud is the same core with the automation layered on top and the hosting, updates and backups done for you.