AI-native learning

You learn from AI all day. You remember almost none of it.

Medha pulls key concepts from your AI chats and resurfaces them at the perfect time so they stick. No flashcards. No extra work.

Start Remembering

Works with any AI — browser, IDE, or terminal

See how it works ↓
You: How do transformers work?
The key insight is the attention mechanism. Each token computes a weighted sum of all other tokens' values, allowing the model to focus on relevant parts of the input regardless of distance.
◆ SAVED BY MEDHA
Attention mechanism: each token computes a weighted sum of all values, enabling focus on relevant input regardless of distance.
Auto-Capture

How does Medha capture your learning?

While you chat with any AI, Medha silently extracts learning concepts in the background. Browser or terminal — nothing changes about your workflow.

Attention mechanism computes weighted sumsSaved
Transformers process sequences in parallelReading...
Backpropagation computes gradients efficientlyQueued
3 concepts captured while you chatted
Works where you work

Where does Medha work?

Browser extension for the AI assistants you already use. One-line setup for Cursor, Claude Code, and any MCP-compatible tool. Everything lands in one place.

Browser
ChatGPT Claude Gemini DeepSeek Qwen Kimi NotebookLM Grok
Chrome extension — 8 AI assistants
IDE & Terminal
MCP
Cursor, Claude Code, Windsurf — any MCP client
Smart Feedback

How does Medha evaluate your knowledge?

You answer in your own words. Medha tells you what you nailed, what's missing, and where to focus next. So you always know exactly where you stand.

Your score
70%
Retention after one week
Attention — solid Parallel processing — good Long-range deps — shaky Position encoding — gaps
Progressive Difficulty

How does progressive difficulty work?

Start with recognition. Graduate to recall. Then prove you can explain it. Each level builds real, durable understanding.

01 — Recognize
Pick the right answer
Multiple choice. Low pressure.
02 — Recall
Fill it in yourself
No hints. Pure memory.
03 — Explain
Prove you own it
Open response. Medha evaluates.
Knowledge Map

See what you know. Spot what you don't.

One dial per topic. Concepts you have mastered sit in the core; the ones slipping drift to the rim, so you see what to review at a glance. Free for everyone.

How much does Medha cost?

Free
No card required
  • 50 captures/month
  • Spaced repetition
  • LLM-evaluated feedback
  • Email review reminders
  • Knowledge map
Start for Free
Pro ✦
$5 per month
  • Everything in Free
  • Unlimited captures
  • API access (MCP server)
Get Pro — $5/mo

Frequently asked questions about Medha

Any AI assistant in your browser via our Chrome extension. Any MCP-compatible tool in your IDE or terminal via our MCP server. New tools are added automatically.
Medha only processes the Q&A pairs you explicitly select — whether from the Chrome extension or via our MCP server in tools like Claude Code and Cursor. Your full conversations are never sent. Selected pairs are processed to extract learning concepts, which are stored for your reviews. You can delete your data anytime.
Your concepts stay. You keep everything you've captured. You revert to the 50-capture limit for new captures only.
iOS and Android are coming soon. For now, the review page works great on mobile browsers.
Import is on our roadmap. For now, Medha auto-generates review material from your AI chats — no manual creation needed.
No. Your learning concepts are yours alone. We never use them for model training or sell them to anyone.

Most of what AI taught you this week is already fading. Medha brings it back.

Start Remembering See How It Works