AI Agents Overview
AI agents are configurable coworkers inside your app. You define what each one handles, which tools it can use, and how people reach it — then it takes on work, follows through, and reports back.
What agents can do
Agents are here to do work— not just answer questions. Describe what you need in plain language; the agent reads the thread, uses the tools you've allowed, and follows through. Run as many as you need per app: a support bot that triages tickets, an ops assistant that chases stale deals, a research agent that summarizes before a call — each with its own instructions and permissions.
- Handle conversations — reply in-app, on a record, Slack, Telegram, Gmail, webhooks, or voice. People message in; the agent does the back-and-forth.
- Work in your app — look up and update records, run automations, manage tasks, and remember context from earlier threads. Routine data work without doing every step yourself.
- Follow up on its own — send email or Slack messages, fire webhooks, and run on a heartbeat or reminder schedule. Agents can start work, not only react when someone asks.
- Reach outside your app — search the web, fetch URLs, run code, generate images, and call other agents. When you grant a capability, the agent uses it to finish the task.
Where to start
Creating agents
Open the Agents library, write instructions, pick a model, and set limits.
Starting threads
Hover an agent in the sidebar and click the new-thread button.
Channels
Connect Slack, email, webhooks, heartbeats, and more ways for agents to run.
Data, connections & integrations
Synced tables vs live third-party actions — and where to connect each.
Tools
What each agent can do — data, email, web, integrations, and more.