⌘⇧S
Screenshot solve
Drag any problem on screen and get a step-by-step solution with LaTeX formatting in the side panel.
Beta — free, no credit card
Solve any homework, instantly.
Screenshot a problem or auto-solve an entire quiz page — right in your browser, powered by frontier models: Claude Opus 4.7, GPT-5.5, and Gemini 2.5.
No credit card. No login. Free during beta.
Two ways to solve.
Math Agent is the only extension that combines an on-screen capture tool with whole-page autosolve.
⌘⇧Q
Auto-solve a quiz
Press once on Canvas, Blackboard, or any quiz page — Math Agent detects every question and fills the answers.
Everything you need to power through homework.
One extension. Every problem type. Built to stay out of your way.
Education modes
Switch between Detailed, Brief, and Tutor — pick how the AI explains.
Not just math
Works for physics, chemistry, biology, and statistics homework too.
LaTeX formatted
Formulas render cleanly with KaTeX so derivations stay readable.
Follow-up chat
Ask follow-ups in the side panel — the conversation keeps full context.
Keyboard shortcuts
⌘⇧S to capture, ⌘⇧Q to autosolve a page, ⌘⇧X to pick one question.
Light & dark
Side panel adapts to your system theme automatically.
From homework to answer in three steps.
- 1
Install
Add Math Agent to Chrome and pin it to your toolbar.
- 2
Capture
Press ⌘⇧S to screenshot a problem, or ⌘⇧Q on a quiz page.
- 3
Solve
A step-by-step answer appears in the side panel. Ask follow-ups.
Frequently asked questions.
Is it free?
Yes — the free tier allows 10 solves per day. Pro removes the cap and is launching soon.
Which subjects does it cover?
Math, physics, chemistry, biology, and statistics. Anything a vision-capable LLM can read off a screenshot.
Does it work on Canvas / Blackboard / McGraw Hill?
Yes. Auto-solve detects generic radio, checkbox, and free-text question groups, so most LMSes work out of the box without per-site adapters.
What model powers the answers?
We route to the best frontier multimodal model for each task — Claude Opus 4.7 by default (best for hard, multi-step problems), with GPT-5.5 and Gemini 2.5 Pro available too. For speed/cost-sensitive solves, lighter models like Claude Haiku 4.5 and GPT-5.5 mini are one toggle away. The backend is OpenAI-compatible, so power users can point at their own endpoint.
Do you store my screenshots?
No. The backend is stateless — screenshots are forwarded to the model and discarded. Nothing is persisted server-side.