ANVE: the private AI office, on the record. See what that means

ANVE

ANVE maker · updated 2026-10-16

Your agents, on your own models.
An OpenAI-compatible API inside the office network.

Short answer. The models on an ANVE unit are served by a runtime (Ollama or vLLM) that already speaks the OpenAI request format, so tools written for the OpenAI SDKs, agent frameworks such as OpenClaw and Hermes, and coding agents can change one base URL and use them instead of a cloud endpoint. Prompts and outputs stay in the building and there is no per-token bill. What is not on ANVE yet is the governed ANVE API on top: a key per user or agent, and every request in the same audit trail as people.

01

What changes in your code

Usually two settings: the base URL (the appliance address on your network) and an API key. Chat-completion style calls stay the same. Per-user keys issued and audited by ANVE arrive with the governed API, which is not on ANVE yet.

02

Fit by tier

WorkloadFit
Summaries, extraction, classificationEvery tier
Agents with tool use, longer reasoningANVE Office and Office Pro (27B to 35B-class)
Many agents at once, strong code generationANVE Office Pro; we say so on the box
03

Status today

Part of the ANVE suite design. It has not yet been shown running in a working setting, and we will say so until it has.

04

Questions people ask

What is an OpenAI-compatible API?
An HTTP API that accepts the same request and response format as OpenAI’s, so existing SDKs and tools work by changing the base URL. See the glossary entry.
Can coding agents use a local model?
Yes, if the model is strong enough. Small 4B-class models suit extraction and summaries; coding agents need 27B-class models or larger. ANVE runs 27B-class by default, and heavy coding use belongs on Office Pro.
Does agent traffic leave the office?
No. The endpoint runs on the appliance on your network. Logging each request against a named user arrives with the governed ANVE API, which is not on ANVE yet.
05

What is not real yet

No sealed ANVE unit exists; every image of the appliance on this site is a labelled render. The software runs today on a development machine. No customers yet, no certifications, no delivery date, and the final price is not announced.

Reservations are ₹999 in India or US$12 elsewhere, refundable in full until you confirm an order in writing (terms). Offices that cannot use cloud AI can apply to be one of ten design partners.