Engram
Engram documentation
Engram turns your handwritten notes — a OneNote export, or any PDF — into a local, Obsidian-native knowledge base. It's an organizer for what you've already learned: searchable summaries, connected into a concept graph, that show you the shape of what you know.
What Engram is
An index and connector, not a transcriber. For each note it writes a concise, searchable summary of the page's key ideas and all its formulas, keeps the original page right below as the source of truth, links related notes, and distils a multi-layer concept graph. Everything runs on your Mac (Claude Code or a local model), and edits are durable and incremental.
The toolbar — start here
Every action in the app lives in the top toolbar. Click an icon below to read how it works and why it's designed that way.
Import a course PDF, or add any single PDF as one note.
Read more →Right-click to delete, move, rename, duplicate, or re-transcribe.
Read more →Cross-link related notes across the whole domain.
Read more →Build the multi-layer concept graph.
Read more →Explore the interactive graph; drill Areas → Themes.
Read more →Ask your notes — grounded and cited.
Read more →Switch light / dark / system.
Read more →Models, filters, and the Claude API key.
Read more →Reload domains and notes from the vault.
Read more →Install & requirements
Python 3.11+ and uv on macOS 14+ (Apple Silicon). The first local-model run downloads ~5 GB.
# set up (add --extra local for the on-device model)
uv sync --extra dev --extra local
# build & open the macOS app
cd macapp && ./build_app.sh && open Engram.appRecommended start ★
Try one section group first, check the summaries, pick your model and filters, then scale. See Add.
- Add one course PDF to a domain (see Using Engram → Add).
- Link the domain, then build Concepts (see Connecting knowledge).
- Open the Graph and Chat with your notes.
Engram is an early v0.2 research prototype · source on GitHub · back to home