Docs menu

Agents and the API

Use it

Ask and Agent

The two ways to talk to your brain in Noetix. Ask only reads; Agent can also write pages and record decisions, and asks before anything it can't undo.

The chat on your dashboard has two modes. Switch between them on any message, in the same conversation.

Ask

Ask only reads. It searches your brain, reads pages, follows your written procedures, and answers with sources. It cannot change anything, however it is asked.

Use Ask for questions: "What's our refund policy?", "Who decided to move to Postgres, and why?", "What changed this week?"

Agent

Agent can do everything Ask does, and also write:

  • edit a page, or start a new one
  • record a decision on its own page
  • add a one-off fact to the Notes page
  • remove something from the brain
  • write or change a skill

Page edits happen straight away, because every version of a page is kept and can be restored. Removing something, and adding or changing a skill, always wait for your yes first: Agent shows what it wants to do, and you approve or skip it.

When you only mention something

Chat doesn't fill your brain on its own. In Agent mode, if you mention something worth keeping that the brain doesn't know yet ("we moved the launch to May"), the assistant offers to save it, for example Add to Notes, with Save and Skip. Nothing is saved until you press Save. Ask never saves anything.

Choosing the model

The model menu is next to the send button. Auto picks the model for you. On the early access allowance only Auto runs; once your workspace adds its own AI key (AI and usage), the other models an admin switched on are available.

Your conversations

  • Rename, Save as Markdown and Delete are in the menu on a conversation's title. Deleting a conversation never touches your brain.
  • Leaving doesn't stop an answer. It keeps being written, and opening the conversation again picks it up.
  • Voice input: press the microphone next to Send and talk. The words appear as you speak, and nothing is sent until you press Send. It uses your browser's own speech recognition, so the audio goes to your browser's maker (Google for Chrome, Apple for Safari, Microsoft for Edge). Firefox has none, so the microphone doesn't show there.