On this page

Knowledge Base

Shared sources and per-agent instructions for Carrie and Brandi.

Knowledge → Knowledge Base (/knowledge-base) is titled AI Knowledge Base. Kicker Knowledge. Description: “Configure the shared knowledge and per-agent instructions for Carrie and Brandi.” Loader: Loading knowledge… / Loading location….

What it is

This is the shared library Carrie and Brandi read when they answer. You add Text, Website, or File sources, enable or remove them, and set per-agent instructions plus the chat embed for that location.

It is not Inbox knowledge (/inbox-settings/knowledge). Inbox Pro has its own Knowledge studio. Notes added there are not added here, and the reverse is also true.

Carrie is the communications agent. Brandi is the reputation agent. Both read this library.

Open the screen

  1. Click Knowledge Base under Knowledge.
  2. Use Viewing settings for and the Select Location dropdown. That picker is local to this screen.

This screen is not behind an add-on gate.

What the page shows

Knowledge Q&A — “Custom Q&A pairs all AI agents can use when answering customer questions.” Each row: title, body, {done}/{total} Page(s) Synced, a toggle, and Remove (when you can delete). Pager: Page {n} of {m} · Previous · Next. Button Add new knowledge.

Live empty (no assistant): No AI assistant found for this location.

Per-Agent Instructions — “Customize how Carrie answers on web chat for {location}.” Chip Web Chat Agent. Fields:

  • Agent Name
  • Website URL — “Where this chat widget is installed.”
  • Chat Widget Embed Code with Copy / Copied
  • Agent Instructions with {n}/10000

Empty instructions: Select a location with an AI assistant to edit per-agent instructions. Footer Save Changes / Saving….

Controls

Add Knowledge modal titles: Add Knowledge → Text / Website / File.

Cards:

  • Text — FAQs or house notes Carrie can quote.
  • Website — Sync pages from your website.
  • File — Menus, wine lists, and PDFs.

Then ← Back, field Name (placeholder e.g. Corkage policy). File: File hint UI only — no upload. placeholder wine-list.pdf. Otherwise URL (https://harborgrill.com/happy-hour) or Content (Carrie should say…). Button Add.

On a source row: toggle to enable or disable, Remove if canDelete.

Typical workflow

  1. Pick the location.
  2. Add new knowledge → Text for a policy, Website for a menu URL, File for a PDF name.
  3. Leave the source enabled. Check Page(s) Synced.
  4. Fill Agent Name, Website URL, and Agent Instructions. Copy the embed if this location’s chat should use that knowledge.
  5. Save Changes.

Empty, locked, or preview-only

File upload is UI-only in this app (UI only — no upload.). If there is no assistant for the location, you cannot add Q&A or edit per-agent instructions until one exists.

Was this guide useful?