> For the complete documentation index, see [llms.txt](https://doc.aissist.io/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://doc.aissist.io/tutorial/quick-start.md).

# Quick Start

The fastest path to a working Aissist AI agent - sign up, onboard, build the workspace, test it in Try, deploy it - with a pre-launch checklist.

*Last updated: November 8, 2026*

The shortest path from sign-up to a live AI agent. Each step is short and links to its full page.

{% stepper %}
{% step %}

### Sign up

Sign up at [console.aissist.io](https://console.aissist.io/). The first time, Aissist asks what you want to do first: **Analyse my current support** or **Create my first workspace**. Both start from your business domain.
{% endstep %}

{% step %}

### Onboard

Enter your website's domain and choose a path:

* **Analyse my current support**: connect your help desk (read-only) and get a report of your volume, topics, first contact resolution, handoffs and complaints, with what Aissist would resolve.
* **Create a workspace**: Aissist drafts the agent from your website, and from your help desk if you connect it.

You can go from the analysis straight to a workspace with **Create a workspace from this analysis**. See [Onboarding Analysis](/tutorial/onboarding-analysis.md).
{% endstep %}

{% step %}

### Review the workspace

Review the draft and click **Create workspace**. Then check **Workspace → Setting** (language, channel) and **Character** (identity, purpose, style). Keep the **Instruction** tab short: scenario rules belong in sub-agents.

See [Create a Workspace](/tutorial/create-workspace.md).
{% endstep %}

{% step %}

### Add assets

Go to **Workspace → Assets → Add Asset** and paste a link to your website, a page or a Google file. Then check retrieval with **Debug**.

See [Add Assets](/tutorial/add-assets.md).
{% endstep %}

{% step %}

### Add actions

If the agent needs live data or must change something, such as an order lookup or a cancellation, connect the system in **Connectors**, then add the action in **Workspace → Actions → Add Action**.

See [Add Actions](/tutorial/add-actions.md).
{% endstep %}

{% step %}

### Add sub-agents

Create one sub-agent per scenario, such as order tracking, returns or billing, in **Workspace → Sub Agents → Add Sub Agent**. Link only the assets and actions that scenario needs.

See [Add Sub Agents](/tutorial/add-sub-agents.md).
{% endstep %}

{% step %}

### Try it

Open **Try**, click **Start**, and send the messages your customers send. Read the reply and what shaped it: sub-agents, actions, sources and escalation.

See [Try](/tutorial/try.md).
{% endstep %}

{% step %}

### Deploy

Create a gateway in **Deploy → Gateways** for your help desk. Start with one inbox or group.

See [Deploy](/tutorial/deploy.md).
{% endstep %}
{% endstepper %}

### Pre-launch checklist

Run through this before you send live traffic.

**Instructions and context**

* Global instructions are short and direct; workflow-specific rules are in sub-agents.
* Shared business facts (contact details, opening hours) are on the **Context** tab.
* Facts, tone and workflow logic are not mixed in one long instruction.

**Assets**

* The sources each workflow needs are added; outdated or conflicting content is removed.
* Google Docs, Sheets and Slides are added through a **Google Drive** asset, and shared with the account that signed in to the Drive connector.
* Real questions return the right passages in the asset debugger.

**Actions**

* Each action has a clear name, a trigger scenario that says when it runs, and clear parameter descriptions.
* Each action returns the right data in the action debugger.

**Sub-agents**

* Each sub-agent covers one clear scenario.
* It links only the assets and actions that scenario needs.
* **Control** settings (no response, close) are on only where you mean them.

**Testing**

* Common questions, edge cases and sensitive requests are tested in **Try**.
* Workflows that depend on the customer use realistic **Input Context**, such as an email or order number.
* Each issue is fixed at its source, then the same scenario is tested again.

**Deployment**

* The gateway manages one narrow area first, such as one inbox or group.
* Escalation and handover rules send each case to the right team.
* Someone reviews live conversations daily for the first days before coverage grows.

### Do it with Ask Engine

Most build steps can start from a sentence. The create pages for sub-agents, assets and actions are Ask Engine: you describe what you want, Ask Engine drafts it from what your workspace already knows, and you review it before anything is created.

> Questions about delivery times and where an order has got to.

> Look up an order's shipping status by order number when a customer asks where their parcel is

Once live, ask about the results:

> List all the primary reasons for escalation for recent conversations, and identify opportunities to reduce AI escalation.

See [Ask Engine](/capabilities/ask.md).

### Best practice

* Build one repeatable workflow first. Test it, deploy it, monitor it, then add the next.
* Start live traffic small and grow it as results hold up.
* Fix knowledge before tuning instructions: most wrong answers come from missing or conflicting content.

### Related

* [Tutorial](/tutorial.md)
* [Instructions and Information](/capabilities/instructions-and-information.md)
* [Success Metrics](/success-metrics.md)