Overview
Build a specialist that works for your business around the clock: a support rep, a bookkeeper, or a follow-up assistant. In this video, you build a custom AI agent. The agent reasons over your projects, calls tools, and remembers past conversations.
TL;DR: A Taskade AI agent combines a custom system prompt, built-in tools, and persistent memory. Give it knowledge sources and slash commands, test it in chat, then share it with your team or embed it on a public page. Watch the walkthrough, then create an agent.
What you will learn
- How to define an agent's role, tone, and goals
- How to add tools so the agent can take action, and not only chat
- How persistent memory keeps context across sessions
- How to share the agent with teammates or embed it on a public page
What an agent can do for you
Each capability maps to a real job your agent can run without you:
| Agent capability | What it does for your business |
|---|---|
| Custom tools | Book a job, update a CRM, or send an invoice |
| Persistent memory | Remembers a client's history across chats |
| Knowledge sources | Answers from your own docs and files |
| Slash commands | Runs a repeatable task on command |
| Public embedding | Puts a support agent on your website |
Quick setup
- Open Agents in your workspace.
- Click New agent and pick a frontier model.
- Write a system prompt that describes the agent's job.
- Attach knowledge sources, tools, and slash commands.
- Test the agent in chat, then publish or share.
If a reply is not right, click Regenerate on the agent's latest reply. You do not need to retype your message. Then tune the prompt until the answer is right.
Go deeper
These walkthroughs show how to train AI agents on your own files. They also show several agents at work together.
Train AI agents with any file, then edit apps visually.
Multi-agent orchestration: let agents collaborate on bigger jobs.