AI Agents

Most of what Scholarly creates isn't a single answer from a chatbot — it's the work of an AI agent. When you ask for a video lecture, a study guide, or a set of flashcards, an autonomous agent takes that goal and completes it end to end: it reads your material, plans the work, uses the right AI models, researches the web when needed, runs code in its own sandbox, and hands back a finished, cited artifact.

This page explains what that means and why it matters for the quality of your study materials.

Agent vs. chatbot

A chatbot answers one message at a time and waits for your next prompt. An agent is different:

• It takes a goal ("turn these three PDFs into a 10-minute lecture"), not just a question.
• It plans the task into steps and decides which tools and models each step needs.
• It uses tools — web search, code execution, file rendering — not just text.
• It checks its own work against your sources and fixes mistakes before finishing.

You don't have to prompt each step. You pick a tool, add your sources, and the agent runs the whole job.

What every Scholarly agent can do

• Use multiple models. Each step is routed to the model that does it best — strong reasoning models for planning, writing models for prose, and dedicated models for vision and speech. See Choosing an AI Model.
• Run code in its own sandbox. Every agent gets a private compute environment where it can transform data, build tables, and render files like slides, audio, and video.
• Research the web. When your uploaded sources leave a gap, the agent searches the web, reads what it finds, and cites it alongside your own material.
• Stay grounded in your sources. Agents start from the PDFs, notes, slides, and recordings you upload, so outputs reflect your actual course — not the open internet.
• Self-review. The agent re-checks the result against your material and attaches citations before saving it to your workspace.

How an agent works, step by step

1. Reads your sources — ingests and understands the files you uploaded.
2. Plans and researches — outlines the work, picks models, and fills gaps with web research.
3. Runs code in its sandbox — computes, formats, and renders the artifact.
4. Delivers and cites — self-reviews against your sources, adds citations, and saves the finished result.

Which tools are agents?

Most AI create options are agent-powered, including AI Video Lectures, AI Slides, AI Podcasts, Deep Research, Quizzes & Exams, Study Guides, Worksheets, AI Spreadsheets, Infographics, Mind Maps, Story Books, and AI Flashcards. In the New menu and on the home page, these are marked with an AI Agent tag.

A few tools are lighter "content" surfaces you work in directly — like the AI Essay Writer, AI Chat, and the AI Image Playground. These are marked with a Content tag instead.

We build the agents for you

Scholarly is not a build-your-own-bot platform. We design, tune, and maintain the agents so that every student gets expert-grade study materials on the first try — no prompt engineering or setup required. You bring your material, pick a tool, and the right agent does the work.

To see the bigger picture, visit the AI Agents overview.