Use case
Upload documents and get structured intelligence: chunked, embedded, entity-extracted and connected. Ask questions in natural language and get cited answers from your own corpus.
What it does
Memo is document AI you can actually use. Upload a corpus and get back something structured: chunked and embedded text for semantic search, an entity-and-relationship graph you can explore, and a question-answering layer that cites its sources. No prompt engineering, no pipeline to build — it runs on every document you file.
01
Chunk
Documents are converted to text and split into passages — markdown, HTML and text natively, PDFs via PyMuPDF.
02
Embed
Each passage becomes a vector with OpenAI text-embedding-3-small, so search matches on meaning.
03
Extract
A language model pulls out entities and typed relationships from every passage.
04
Connect
Entities resolve to canonical nodes, so the same concept across documents is a single point in your graph.
Two ways to query
Semantic search
Find passages by what they mean, not the exact words. Filter by document, type or date and get back the most relevant chunks across your whole library.
Ask, with citations
Pose a question in plain language. Memo retrieves the right passages and composes an answer that points back to the documents it came from.
Papers, reports and notes become one connected, askable graph instead of a folder of PDFs.
Make handbooks, contracts and SOPs searchable by meaning, with cited answers your team can trust.
Years of documents, finally findable — and connected by the people and topics inside them.
Developers can query the same engine over the HTTP API or connect an assistant over MCP.
Upload your first document and watch a knowledge graph build itself. Free to start; Pro when your library grows.