Answers From Your Own Content, Drafted by Claude
Anthropic serves the Claude model family over a hosted API. Connected to AI Intelligence Hub, Claude becomes the model behind any agent you assign it to. It reads the passages retrieved from your library and writes the answer, with a citation to each source. Your video, audio, images and documents stay in Nexus under your own access rules. Claude receives the question and the retrieved text, never the library.
What you can do together
- Ask AI Intelligence Hub a plain-language question and get an answer grounded in your own recordings and documents, with a citation card for every source it used.
- Run a different model per agent: Claude for the policy library, a self-hosted model for material that must not leave the site, each agent with its own prompt, scope and tools.
- Scope an agent to a folder, a category, a set of tags or an attribute value, so it answers from that subset and never from anything the asking user could not open themselves.
- Put Claude inside a workflow node, so a multi-step workflow with a human approval step runs on the same model as the chat.
- Attach a file in the conversation for analysis and carry the context through follow-up questions.
How it connects
AI Intelligence Hub calls Anthropic's REST API with your organization's own account and key. The provider is configuration, set for the deployment or for a single agent, so one agent can run Claude while another runs a different provider at the same time.
When a user asks a question, AI Intelligence Hub retrieves matching passages from the transcripts, OCR text, visual descriptions and document text that user is permitted to open. The retrieved passages and the question go to Claude as one prompt. The answer comes back and is shown with citation cards that open the source at the cited moment or page.
Embeddings for retrieval are generated by a separately configured embedding provider, hosted or self-hosted. Anthropic sees prompts and answers, not the index and not the media.
A scenario
- SetupIT enters the city's Anthropic key in AI Intelligence Hub, picks Claude as the model for the Council Records agent and scopes the agent to the meetings and ordinances folders.
- 09:10A records analyst asks, "When did council last discuss the sidewalk vending ordinance, and what was decided?" The agent retrieves passages from the meeting transcripts and staff reports she can open and sends them to Claude with the question.
- Seconds laterThe answer names the meeting and the motion, citing two transcript moments and one staff report page. She opens a citation card and the recording plays from that moment in Nexus.
- 09:14She follows up: "Draft a short summary of that decision for the newsletter." The conversation carries the context, and Claude writes it from the same cited passages.
- Same morningA council aide with narrower permissions asks the same agent. The closed-session recording is outside his access list, so it is never a candidate for retrieval. The prompt and passages went to Anthropic; the recordings and documents never left the city's deployment.
What stays where
Anthropic keeps serving the model
The account, the key, the usage and the choice of model per agent are yours.
AI Intelligence Hub sends prompts, not the library
The question and the retrieved passages travel to Anthropic's API and the answer travels back. Embeddings run through the embedding provider you configure, and the media never moves.
Nothing is replaced
Switching an agent to another provider is a configuration change, and the workflows behind it do not change.
Where processing runs
Claude runs in Anthropic's hosted service, so the prompt is processed there. The platform itself runs as SaaS, in your own cloud subscription, on your own servers, or air-gapped. An air-gapped deployment cannot reach a hosted API and serves its language models through Ollama or vLLM instead.
Products and solutions
Next step
See it on your own Anthropic Claude instance.
We will show the connection made, the data moving and the output, then size it for your deployment.
Contact VIDIZMO
sales@vidizmo.ai
+1 571-969-2180
vidizmo.ai