Skip to main content

​
What is a chat?

A chat is a conversation with the Attensira agent about your own workspace. You ask in plain language — “which prompts lost citations this month?” — and the agent queries your tracked data and answers with the numbers behind it. Each chat is a session you can reopen and continue. Chats are scoped to one Workspace. The agent answers from that workspace’s Prompts, Topics, Models, Answers, and citations, and cannot read another workspace’s data.

​
What can the agent reach?

The agent reads the same measurements the dashboard shows: your Prompts and the Topics they sit in, the Models tracked for the workspace, the Answers those models returned, the Mentions and Citations inside them, and the Sources those citations roll up into. It can also act — adding prompts, adding a competitor, and creating or running Automations. It cannot reach anything you have not tracked. If a Model is not configured for the workspace it was never queried, so there is no data to summarise; the agent should say that rather than report zero. If a Prompt does not exist, no answer about it exists either.
The agent inherits the measurement rules, not just the data. Null means not measured, a measured zero means runs happened and found nothing, and a delta below the 95% significance threshold means no change we can prove. Ask the agent for n when a number surprises you.

​
What is an “agent model”, and how is it different from a Model?

These are two different things that both get called “model” in casual speech, so Attensira keeps them apart:
  • A Model is a surface you track — ChatGPT, Perplexity, Claude, Gemini, Google AI. It is the thing being measured. Each has a stable channel id that survives the underlying model being swapped out.
  • An agent model is the LLM that runs your chat. It is the thing doing the measuring’s explanation. Switching it changes how the agent reasons and writes, and changes nothing about your data.
You pick the agent model from the selector in the chat. A different agent model may phrase an answer differently or take a different route through your data, but the underlying measurements are identical — the numbers do not move because you switched.
Switching agent models is not a way to get a second opinion on your visibility. Every agent model reads the same stored runs. If a number looks wrong, check n and the tracked flags, not the agent model.

​
Auto, and why it is the default

Auto is the default and is right for almost every question. It routes each turn to whichever model clears the bar for that question, across every provider on the list — so a short lookup runs somewhere cheap and a hard analysis runs somewhere capable. In practice your Credits go further under Auto than under any single model you could pin. Overriding is fully supported. The selector also offers named models from Anthropic, OpenAI, Google and DeepSeek; pick one when you want a specific model’s voice, or when you are comparing how two of them write the same brief.

​
Effort is chosen for you

Each model in the list carries a fixed effort, shown as a tag beside its name — High on the thorough ones, Fast on the quick ones. There is no separate effort control. This is deliberate. Effort and model are not independent choices: a fast, cheap model told to think as hard as it can is slower and worse than simply picking a stronger model. Pairing them removes a combination that never had a good outcome, and leaves you one decision instead of two.
Under Auto, the model that served your turn and the model you selected are not the same thing. Billing always follows what the call actually cost, never the id that was sent.

​
Asking a good question

1

Name the window

“Last 30 days” or “since the start of the quarter”. Without a window the agent picks one, and short windows at low sampling depth are mostly noise.
2

Name the scope

A Topic, a Prompt, a Model, or a competitor. Broad questions return broad answers.
3

Ask for the denominator

Rates share a denominator across brands, so competitor rates do not sum to 1. Ask for the run counts if you plan to quote the figure.
4

Follow up in the same chat

The session keeps context, so you can narrow without restating everything.

​
Costs and limits

Asking the agent spends Credits, because each question runs a model and may run tools on your behalf. A long analytical question costs more than a short lookup. A chat that triggers an Automation run spends credits for that run as well. Some questions take long enough that the agent returns a running status and finishes in the background; reopen the session to read the completed answer. The same behaviour is exposed over MCP through ask and get_session.

Automations

Turn a question you keep asking into a scheduled instruction.

Knowledge and skills

Shape how the agent writes for your brand.