Getting Started

Set up your workspace, connect core tools, and launch your first AI agent.

Set Up Your First AI Agent

Start with one focused agent and one safe test channel. A narrow first rollout is easier to review, improve, and trust.

Before You Click Create

Confirm these pieces first:

shared context is present in AI Agents > Memory
required tools are connected in Settings > Integrations
the agent has a channel where the team can review its work
any plan-gated tools are available in Settings > Billing
the task can be tested without affecting real customers

1. Open The Agent Area

Open AI Agents > Agents. This is where you create and manage the AI workforce for the workspace.

The AI Agents hub also includes:

Memory for shared business context, knowledge, and long-term memory
Skills for agent capabilities
Voice for call-related setup when available

2. Choose A Focused Starting Point

Pick a role that has a clear outcome. Good first agents include:

an internal operations assistant
a marketing research assistant
a Google Business Profile assistant
a contact cleanup assistant
a campaign drafting assistant

Avoid starting with an agent that owns a full customer journey until you have tested its behavior.

3. Configure The Agent

For the first pass, keep configuration simple:

name: make the job obvious to teammates
description: explain what the agent should and should not do
goals: define what success means for this agent
skills: enable only the capabilities it needs
integrations: connect the accounts it will actually use
approvals: require review for sensitive write actions

If the agent needs company-wide context, put that context in AI Agents > Memory instead of repeating it on every agent.

4. Put The Agent In A Real Channel

Use a channel that matches the work:

an operations channel for internal tasks
a marketing channel for campaign work
a customer or project channel for account-specific work
a test channel for early evaluation

Tell the agent what you expect, who should review the output, and what actions need approval.

5. Test With A Bounded Task

Start with a prompt that has a clear finish line.

Example:

Review the business profile and connected Google Business Profile context.
Draft three weekly post ideas for our Denver HVAC service area.
Do not publish anything. Return the drafts for review.

Review whether the agent used the right context, respected approval boundaries, and produced work that a teammate can act on.

6. Learn The Supporting Tabs

Memory stores shared business context, knowledge, guidelines, and longer-term memories.
Skills controls the capabilities agents can use.
Voice groups call history, numbers, A2P setup, templates, the playground, and voice settings.
Agent History helps admins inspect past agent activity.

7. Add Proactive Behavior Carefully

Use proactive behavior only after a manual workflow has been tested. For recurring checks, consider Proactive Heartbeats. For structured trigger/action flows, use Workflows & Automation.

Next Steps

Using the Help Center

AI Agents