What are AI agents? Guide

An AI agent is a configurable worker you set up once and then rely on to carry out a specific job. Where the assistant helps you in a live conversation, an agent is built around a purpose — monitoring a brand, writing blog posts, drafting replies — and equipped with the tools, skills, and knowledge it needs to do that work well.

Think of each agent as a specialist you hire and brief. You describe what you want, connect the accounts it should use, and it runs the task and reports back.

What makes up an agent

Every agent is defined by a set of capabilities. Together they decide what the agent knows, what it can do, and how it keeps you informed. You will find these grouped under Configure on each agent's detail page.

Tools and integrations

Tools are the actions an agent can take. Built-in tools are always available — things like emailing you, saving results, searching the web, working with files, and generating images. Integrations extend that reach to your connected accounts, such as Gmail, Slack, or Shopify, and show which account each one uses.

Note: An agent with no explicit tools set can still use all the built-in tools. Integrations, by contrast, need the matching account connected before the agent can use them.

Skills

Skills are packaged expertise — proven, step-by-step playbooks for a specific job, like writing an SEO blog or monitoring a brand. Installing a skill gives your agent both the method and the means: it also grants the extra tools that skill relies on, so you do not have to add them one by one.

Knowledge

The knowledge base is everything the agent reads on every run — product docs, brand guidelines, price lists, FAQs — plus facts it picks up as it works. Grounding an agent in your specifics is what keeps its output accurate and on-brand rather than generic.

Files

The Files area holds two things: the agent's own persona and memory files (who it is, its purpose, and what it has learned) and every output file it produces, such as PDFs, images, and reports. You can open, preview, and download any of them.

Secrets

Secrets are a secure vault for the API keys and tokens an agent needs to call outside services. Values are encrypted, hidden after saving, and referenced by name — so a sensitive credential never sits in plain sight.

Channels and notifications

Channels control how the agent keeps you posted: email a summary, ping you on WhatsApp, or stay quiet until there is something worth sharing. You set this up in plain language so the agent reaches you where you already work.

How you set one up

Most of the configuration above is handled for you by the AI builder, Stan, on the agent's Build tab. You describe the agent in plain English and it wires up the right tools, skills, knowledge, and notifications as it goes.

1. Open the Build tab and describe what you want the agent to do.
2. Connect any accounts it should use when the builder asks for them.
3. Review the capabilities on the Tools, Skills, Files, Secrets, and Knowledge tabs.
4. Do a test run to confirm the agent behaves the way you expect.

You can manage Knowledge and Secrets yourself at any time — for example, uploading a new price list or pasting in an API key. For most other changes, the quickest path is simply to ask the builder.

Tip: Agent and output files are read-only for you. To change how an agent thinks or what it produces, ask the builder rather than editing files directly.

Staying in control

Two safeguards help you make changes with confidence. A test run lets you see exactly what an agent will do before you rely on it, and version history lets you roll back if a change does not work out. Between them, you can experiment freely and always return to a known-good setup.

Agents vs. the assistant

Use the assistant for live, back-and-forth help in the moment. Reach for an agent when you have a repeatable job you want handled consistently — with its own tools, its own knowledge, and its own way of reporting back to you.