Most of the work of an answer is finding the right passages, and that runs in your own deployment. A model is called only when an answer needs explaining, and then a small one.
Octy runs wherever your engine runs: Atlantic-hosted, your AWS, Azure or Google Cloud account, your own servers, or air-gapped on local models.
You decide who can ask, from which sites and networks, how often, and the most it may spend in a month. Octy answers and explains; it never approves or pays.
How Octy answers.
Every question follows the same path. The controls come first, so a question that should not be answered costs nothing, and the search runs before any model is considered.
- A questionFrom the website, your intranet, Teams, Slack, another agent through MCP, or your own application.
- ControlsSign-in, the sites and networks allowed, rate limits and the spend cap, checked before anything else.
- RetrievalA search of that profile's knowledge only, within the access of the person asking, inside your deployment.
- EvidenceThe answer arrives with its sources, checked against them before it is sent.
Low cost by design.
Sending every question to a large model, with whole documents as context, is the expensive way to answer. Octy searches first, in your deployment, and calls a model only when an answer needs explaining. An explained answer is a short prompt of a few passages, run on a small model, and typically costs a fraction of a cent, estimated.
You pay your own provider
Octy calls models through your own gateway, on your own provider account or on local models. Atlantic resells nothing.
A cap on every profile
A monthly spend cap, enforced by the gateway before a model is called, and a monthly limit for Octy in LightHouse™.
Cost you can see Coming
LightHouse shows each profile's spend against its cap, its cost per answer, the share answered from retrieval, and the estimated saving.
Illustrative shape only. Your costs are measured in your own account, from your provider's prices, and every saving is estimated.
Octy, configured for each audience.
Octy runs as profiles: the same code, character and rules, each with its own knowledge and its own limits. A public profile on your website, a customer profile behind sign-in and a staff profile on your intranet can run side by side, and none can read another's knowledge.
- KnowledgeYour documents, connected sources such as SharePoint, Confluence, Amazon S3 and Google Drive, or published regulation.
- AudienceThe public, signed-in customers, or your staff through your own identity provider.
- Disclosure tierPublic, NDA or internal. A public Octy holds only knowledge cleared for publication.
- ChannelsWebsite widget, the LightHouse portal, Microsoft Teams, Slack, MCP and the API.
- ControlsRate limits, allowed sites, IP allow and deny lists, allowed pages, a bot check and a spend cap.
- TrustCitations, the evidence panel, signed answers, audit, and insight into the questions people ask.
The boundary around each profile's knowledge is enforced in the search itself, so the wording of a question cannot widen it. Profiles managed in LightHouse, with public-use controls and signed answers, are coming; the documentation marks each feature available or coming.
In your own cloud, or fully offline.
Octy runs inside your deployment of the Atlantic Agentic Engine, beside your gateway and your indexes. Your documents, questions and answers stay in your environment; LightHouse receives counts, timings and costs.
- Atlantic-hostedA dedicated account, run for you, in the region you choose.
- AWSIn your own account, on Amazon Bedrock or the Claude API.
- Microsoft AzureIn your own subscription, on Azure AI Foundry.
- Google CloudIn your own project, on Vertex AI.
- Your own serversDocker or Podman, in your data centre.
- Air-gappedNo network path out, on open-weight models running on your own NVIDIA GPUs.
Always-on, ephemeral or scheduled: Octy works under every deployment profile, so a profile used in working hours can sleep outside them.
You control who can ask, from where, and how much it can spend.
Every control is checked before a model is called, and Octy keeps the same rules whatever the profile's configuration.
Who can ask
Sign-in through your own identity provider for staff and customers, with each person's access applied to the search. A bot check for public use.
From where
The websites allowed to embed it, IP allow and deny lists, and the pages on which the widget may appear.
How often
Rate limits for each person or visitor, and for each profile overall.
How much
A spend cap for each profile, enforced by the gateway, and a kill switch outside the agent.
Octy refuses any request to approve, pay, change access or a limit, or send content out. Work that needs doing is handed to an agent through an approval card, and a named person approves it in LightHouse.
Evidence you can verify.
Every answer comes with the passages it was written from. A signed receipt records which Octy answered, when, and from which passages, so anyone holding the receipt can check that the answer they were shown is the one that was given.
Citations
Every answer cites its sources, and is checked against them before it is sent.
Signed answers Coming
Each answer carries a receipt signed by the Octy profile that gave it, verifiable in the browser or from the command line.
Audit and insight
Every answer is recorded, audit-ready, and the questions people ask show where your knowledge has gaps.
Where your people already work.
The same Octy, with the same answers and controls, on each channel you switch on.
- Website widget
- LightHouse portal
- Microsoft Teams
- Slack
- GitHub Copilot, through MCP
- Copilot Studio, through MCP
- API
Put Octy on your knowledge.
Our certified forward-deployed engineers set up and evaluate your first Octy with you, on your own knowledge and in your own cloud. The documentation says what is available today and what is coming.