Local by default
Your decks, cards, assets, and complete review history live on your device.
Local-first learning with AI agents
Use your agent to turn a codebase, research, or any topic into source-grounded recall prompts. You choose what enters your study plan; FSRS helps it stick.
A deliberate capture loop
Whether you are learning a codebase or a broader topic, generated content never becomes a scheduled card by accident. Every prompt passes through an inspectable workflow.
Work through code, documentation, research, or study material with your agent.
The agent proposes concise drafts grounded in the material, with rationale and sources.
Answer it, edit it, and accept only what belongs in your study plan.
FSRS-6 schedules each review to build durable recall efficiently.
Approval, not autopilot
Agent proposals remain local, editable, and unscheduled until you accept them. See why a card matters and where it came from before it joins your reviews.
One review engine
Desktop and terminal reviews use the same Rust StudySession. Card selection, scheduling, daily limits, and history cannot drift apart.
Built to remain yours
Your decks, cards, assets, and complete review history live on your device.
FSRS-6 adapts to your memory with configurable retention and learning steps.
Create source-grounded drafts from codebases, research, and other material with Codex, Claude Code, pi, or your own scripts.
Write naturally with code highlighting, images, GIFs, links, lists, and tags.
Sync through any remote using your existing Git credentials—no stored token.
Track recall, activity, streaks, grade distribution, and deck performance.
Start remembering
Install the desktop app and CLI together on macOS or Linux, or download the package for your platform from GitHub Releases.
View all downloads$ curl -fsSL https://raw.githubusercontent.com/shettyh/rem/main/install.sh | sh
$ npx skills add shettyh/rem --skill rem-card-capture --global
Windows installers and a standalone CLI zip are available from Releases. Current macOS and Windows builds are unsigned; see the launch instructions.
Build durable understanding