Quickstart

This walks you through the onboarding wizard you see right after you sign up. It is seven steps, and you can skip most of them and come back later. The goal is to get your agent answering from your knowledge and receiving product data.

1. Sign up and create your workspace

Create an account at app.airhop.ai. The first step names and brands your workspace. Your workspace gets a slug, which Airhop uses for things like your dedicated support inbound email address ({your-slug}@in.airhop.ai, used for human handover).

2. Invite your team

Add teammates who will help answer escalations and watch accounts. You can skip this and invite people later.

3. Add your knowledge

Point Airhop at the sources it should answer from. You can add several at once. The most common one is your help center URL: Airhop crawls it and indexes the pages so the agent can answer grounded questions. See Connect your knowledge for how crawling works.

This step kicks off the crawl in the background. You do not have to wait for it to finish before moving on; answers improve as indexing completes.

4. Connect Gmail and Slack (optional)

Connect Gmail to learn from past support emails, and Slack to ingest knowledge from your team channels. Both open inline during onboarding. See Integrations for the full list of connectors.

5. Set up human handover

Tell Airhop who to loop in when it is not confident. You set the notify emails here. When the agent's confidence is low, it escalates the conversation to your team by email (and the customer gets a clean "I have passed this to our team" message). See Human handover.

6. Add Airhop to your app

This step gives you two things to install:

1. The support widget, a single script tag you paste before </body> on every page. See Install the widget.
2. The product events SDK, a few lines that identify your users and track what they do. This is what powers activation and retention. See Send product events.

You can copy the snippets yourself, or use the "Have your AI do it" tab to copy one prompt you paste into your AI coding agent (Claude Code, Codex, Cursor, Replit, or Lovable). That prompt installs the widget, wires up product events, and drops a skill file into your repo so future changes keep sending data.

While indexing is still running, this step shows live progress ("Indexing your knowledge... N of M sources ready"). You can install right away; answers get better as the crawl finishes.

7. Go live

The last step confirms your product data is flowing before you head to the dashboard. Once you have added the snippet (or run the AI prompt) and opened your app while logged in, this lights up the moment your first event arrives ("Data is flowing! We've received N events").

You can also ask your agent a question right here to see a live, grounded answer with its citations, before you ever leave onboarding. If the answer is low confidence, you will see a note that the agent would hand it off to a human.

Where to go next