Surfaces
Chat / Embed / Org
Router
ICL-1
Knowledge
Docs + chunks
Atina platform
-> Chat workspace
-> Embed SDK and domain-controlled agents
-> Organization AI training workspace
-> Knowledge ingestion and retrieval
-> ICL-1 capability routing
-> Product feeds and read-only tools
-> Avatar/text response surfaces
-> Usage, reviews, reports, and support
Platform View
Atina is being built as AI assistant infrastructure, not only a single chat page. The platform now supports direct chat, third-party site embedding, organization training workspaces, knowledge ingestion, product feed context, read-only live tools, escalation handling, analytics, support workflows, and avatar-based assistant experiences.Chat Surface
The Chat surface is the user workspace for direct Atina conversations. It supports normal assistant use, conversation history, sharing, ratings, profile controls, plan usage visibility, and, for eligible plans, the Embed SDK setup tab.Embed Surface
The embed surface lets site owners deploy Atina on their own domains with a browser-safe public key, authorized domains, SDK assets, text or avatar mode, recommendations, chat history mounting, latency debug, knowledge sources, product feeds, read-only live tools, and escalation configuration. The public reference is Atina Embed SDK documentation.Organization Surface
The organization surface lets teams create a private workspace, invite members, upload company handbooks, generate lessons, control access, run learner sessions, answer questions from lesson context, and review progress. Start at Organization Dashboard.Routing And Knowledge
ICL-1 adds a dedicated capability-routing layer for embedded agents. Knowledge builders chunk uploaded documents into searchable units. Product feeds give Atina structured public catalog context. Live tools let the assistant perform read-only lookups through approved customer endpoints.Controls And Limits
Plans control daily usage, monthly usage, embed access, organization access, organization member capacity, knowledge base capacity, file size, storage, training modules, AI generation counts, learning sessions, and learner questions. Operational dashboards record conversations, outcomes, escalations, ICL latency, and common questions where available.Safety Boundary
This public infrastructure note describes platform capacity and user-facing architecture. It does not publish private gateway keys, provider keys, internal service URLs, raw deployment credentials, private prompts, or customer-specific secrets.Where To Go Next
Builders should read Embedding Atina on third-party sites and Atina Embed SDK documentation. Team admins should read Organization AI training with company handbooks and open Organization Dashboard.
Related
Continue exploring.
Release note: Embedding, Organization AI, and ICL-1
The July 2026 Atina Labs content refresh documents embed setup, organization training, ICL-1 routing, embedded RAG architecture, infrastructure capacity, and a latency-based speech decision.
Organization AI training with company handbooks
How organization teams use Atina to turn company handbooks and internal documents into private knowledge bases, training modules, lessons, learner support, access control, and reports.
Why Atina rejected the Gemini TTS orchestration hack
An architectural decision note explaining why Atina kept its current avatar response path instead of adding a second Gemini TTS pass that increased user-perceived latency.
