An AI agent that learns with you, not instead of you.
Every Dometrain course now has an agentic assistant built in. On Hands-On courses it edits your files and runs your tests, live in your editor. On System Design courses it works the diagram canvas, placing blocks and connections as you watch. And on video lessons it searches the lecture, jumps the player to the exact moment, and explains what’s on screen.
A real agent, not a chat box.
It plans, acts, and verifies, using the same workspace tools you do. Everything it does is narrated in a terminal beside your code.
Writes and edits your files
Creates new files, refactors existing ones, renames and reorganizes folders, directly in your workspace, with every edit landing as a reviewable line-by-line diff in the terminal.
Runs the exercise tests
It doesn't guess. The agent runs the real test suite, reads the failures, and iterates until things are green, up to eight tool rounds per prompt.
Builds system-design diagrams
On System Design courses the same agent works the architecture canvas: it places blocks, wires connections, and checks your design against the lesson rules, animated live as it builds.
Knows every lecture in the course
On video lessons it searches the transcript of the lesson you’re on or the entire course, quotes the exact moment, and links you straight to it. No more scrubbing for "where was that covered?"
Drives the video player
It seeks to the second something was said, plays, pauses, and changes playback speed on request. Every timestamp it mentions is a clickable link, and it can recap or quiz you on what you’ve watched.
Sees the screen, and marks it up
Ask about anything on screen: it captures the current frame, then annotates the paused video with step-through notes typed live over the frame. /extractcode lifts on-screen code into a copyable block.
Explores SQL databases
On SQL lessons it can run read queries against the in-browser database (PostgreSQL, MySQL, SQL Server, or SQLite) to inspect schemas and data before writing an answer.
Teaches while it types
Flip on explain mode and the agent annotates the code it writes with "why this code?" notes you can hover right in the editor.
Terminal-native
Slash commands like /solve and /clear, @file mentions with autocomplete, and a live status bar showing your weekly allowance as you go.
Hints first. Full solutions only when you ask.
The agent follows a learning-first policy: it nudges you toward the failing line, explains the concept, and asks what you think. It hands over a complete solution only when you explicitly request one.
Describe the problem
Plain English is enough. Reference specific files with @ mentions: “why does @OrderService.cs fail the concurrency test?”
See it work, live
The agent narrates its reasoning and works your actual workspace: editing files, placing diagram blocks, or seeking the video to the moment that answers you. It runs your tests or design checks. Every action is a visible tool line with an expandable result.
Pick your depth
Take the hint and finish it yourself, ask for a deeper explanation, or type /solve for the complete solution with a walkthrough of the approach.
/clear resets the session whenever you want.However you learn, it meets you there.
The stuck learner
You've stared at the same failing test for twenty minutes. The agent runs it, pinpoints the branch you missed, and nudges you there, without spoiling the exercise.
The hands-on builder
You want a senior engineer looking over your shoulder. Ask "is there a more idiomatic way?" mid-exercise, or "review my architecture" mid-diagram, and get a reasoned answer against your actual work.
The time-boxed pro
Evenings are all you have. Let /solve handle the boilerplate steps and spend your energy on the concepts the course is actually teaching.
The rewatcher
"Where did he cover idempotency?" Ask instead of scrubbing. The agent searches every lecture in the course, links you to the exact moment, and can recap or quiz you on what you just watched.
Every account gets an allowance. Bigger plans get bigger ones.
Usage is measured in tokens: everything you send the agent plus everything it writes back. Allowances reset every Monday at 00:00 UTC.
Included with every Dometrain account
- Works on every lesson you can access: video, exercise, or diagram
- Basic model: fast everyday answers
- Up to 10 new prompts per hour
More than 6× the free allowance
- Basic & Lite models: pick per session from the terminal
- No hourly prompt cap
- On every course included in Lite
20× the free allowance
- All models: Basic, Lite & Pro, our most capable
- No hourly prompt cap
- On every course in the catalog
All plans share a light fair-use burst limit so the terminal stays fast for everyone. A single prompt can drive up to 8 tool rounds; /clear starts a fresh session any time. Explain-mode hover notes use extra tokens from your weekly allowance.
Frequently asked
Where do I find the agent?
Will it just do the exercise for me?
/solve and it solves the exercise and walks you through the approach.What does it do on video lessons?
/recap and /quiz. Ask about something on screen and it captures the current frame, explains it, and can annotate the paused video with step-through notes. /extractcode transcribes on-screen code into a copyable block.Do I need Lite or Pro to use it?
What counts against my allowance?
Which AI model powers it?
Can it break my work?
Meet your new pair programmer.
Open any course (video, Hands-On, or System Design), hit a lesson, and ask. It’s already there.