What is anki-api?
An agent-assisted system for generating high-quality Anki flashcards from web content, YouTube transcripts, and documents. Designed for use with coding agents like Claude Code — agents propose cards, you review before adding to Anki.
Highlights
- Agent-assisted, not automated: Agents generate card proposals; you maintain quality control through an interactive review workflow
- EAT 2.0 Framework: Cognitive science-grounded principles (Encoded, Atomic, Timeless) guide flashcard quality through agent judgment
- Multiple content sources: Web scraping (via crawl4ai), YouTube transcripts, DOCX documents
- Full CLI:
ping,decks,models,review,add,quick,find,delete, and more - Web UI: React + FastAPI interface for card review and management
- Simple stack: Python (requests, pydantic, click) — no heavy LLM frameworks
Quick Start
```bash
git clone https://github.com/SystemicVoid/anki-api.git
cd anki-api
uv sync
uv run anki-api ping # requires Anki Desktop + AnkiConnect
```
Requirements
- Python 3.11+ (via uv)
- Anki Desktop with AnkiConnect plugin
- Node.js 20+ & pnpm (optional, for web frontend)
- crawl4ai (optional, for web scraping)
What's Next
This is the foundational release. The system is fully functional for personal flashcard generation workflows. Future work will focus on streamlining the web UI and expanding content source integrations.