ProductUnderstand
Askcoming
Questions about your system, answered with sources or marked unknown
Ask a question about your system. Get the answer with its sources.
Ask is the way into Atlas for anyone in engineering, or a sales engineer on a call: where does this request really go, what changed, which decision governs it. Every answer points at the claims and the evidence behind it, and where nothing has looked, it says so. It is written down and not built yet.
Planned and written down; not built yet. Book a call to hear when it opens.
Questions such as where a request really goes, or what changed last week, cost an engineer an afternoon. An AI chat answers at once and can be wrong with no way to tell.
The model reads the question and Atlas's claims; it never writes a claim. What it answers is drawn from what Atlas can prove.
How it works
- 01
Where it runs
- On the platform, in the console or through your AI assistant over MCP. Planned and not built yet.
- 02
What it reads
- Atlas's claims and the evidence behind each one, so an answer can only rest on what Atlas can prove.
- 03
What you approve
- Which model answers: self-hosted by default, a third-party provider only through a connection you make with your own key.
Each line says how far it is built today. Live means anyone with an account uses it; preview means it runs in early access or as a pre-release; coming means it is written down and not built yet.
What you get
Answers with their sources
Each answer cites the claims it rests on and the evidence behind each claim, so a reader can check it without trusting the model.
comingUnknown, said plainly
Where Atlas has no claim, the answer says so and names what would settle it, without a guess dressed as a fact.
comingModels you choose
Self-hosted models by default; a third-party provider only through your own connection and key, chosen by how sensitive the data is.
coming
How an answer is meant to read: numbered sources you can open, and the part nobody can answer yet said plainly.
Design sketch
Where does a checkout request go after the gateway?
To checkout, which calls ledger over gRPC on port 7443 [1] and payments-db through its pool [2]. Retries to ledger go against ADR-0017 [3].
Sources
- [1]Claim: checkout calls ledger · Envoy route, Kubernetes Service
- [2]Claim: checkout uses payments-db · connection pool config
- [3]Contradiction: ADR-0017 and checkout/src/ledger.rs
Unknown
Whether checkout also calls fraud-check: no claim yet. Traffic from the Envoy access log would settle it.
Answered by a self-hosted model from Atlas's claims
Example data
A sketch of a planned feature; Ask is not built. The question, services and sources are made up.
Every claim on this page has its evidence: the documents that define it, the docs that say how to use it, and the API it runs on.
Proof and links
Documents
- ADR 0021 Model providers and the query surface
- PRD 0001 InOrbit Atlas
Not on the public RFCs host yet, so named here without a link.
API
- RESTcomingNot built yet.
- MCPcomingNot built yet.
The evidence rule: a model saying it succeeded is never evidence. What counts is what was measured, by what, and when.
Your data, your call
Questions and answers are your data: never in a log, and sent to an outside model only through a provider you connected yourself. Billing and terms are in Company, under Your data.
Logs record who did what and the outcome, never content. The platform's own models are self-hosted; a third-party model provider is one you connect with your own key, and it bills you directly, with no line or markup from InOrbit. Separately, our own DPAs with model providers are in progress, none signed yet. InOrbit is not certified: no auditor has attested its controls yet, and the known gaps are listed.
Questions, answered plainly
Can I try it?
Not yet. It needs Atlas served first, and Atlas is in development.
Is it a chatbot over our docs?
No. It answers from Atlas's claims, each with its proof, and says what it does not know.
Will it tell me it is an AI?
Yes, always, as the site's own guide does today.
Next step
See it on your own system
Book 30 minutes with an engineer. We look at your system together and, where it fits, open early access for your team.