Think less.
Remember everything.
Taggard is a memory engine for your notes and documents. Every note gets two kinds of tags — ones you write for browsing, and ones Taggard writes for deep, cross-referenced recall you'll never have to search for by hand.
Two kinds of tags. One for you, one for the machine.
Visible tags are the labels you'd write anyway — a name, a project, a date. Underneath, Taggard builds a second, hidden layer: latent frames, tone, methodology, confidence — signals no one would bother tagging by hand, and ordinary search can't see.
Q3 planning notes
Team agreed to delay the senior hire until Q4 and reallocate budget toward contractor support. Acme renewal still open — revisit after the board review.
Entities resolved: Acme Corp, Q4. Confidence: 86%. Connected to two earlier notes Taggard found on its own.
Everything below runs on its own.
Entities, resolved
People, organizations, places, dates, and products — recognized and matched across every note, so 'Acme' and 'Acme Corp' are the same thing.
Cross-references
Taggard finds when one note builds on, contradicts, cites, or echoes another — and tells you why, with the exact passages compared.
Ask, with citations
Ask a question in plain English. Get an answer built from across your notes — including ones connected by meaning, not just keywords — with sources cited.
Idea evolution
Pick a topic and see it as a timeline: which note built on which, where you changed your mind, and when.
Theme discovery
Notes that share a latent pattern — not just a tag — get clustered automatically, and Taggard can name what connects them.
Knowledge graph
Every note and connection, visualized. Click a dot, follow a line, see your thinking as a map instead of a list.
A real AI stack, not a keyword index with a chatbot bolted on.
Every extraction step — tagging, entity resolution, cross-reference classification, theme naming — is a dedicated model call with its own schema, not one prompt doing everything.
Language models
GPT-class models handle tagging, entity extraction, cross-reference classification, RAG answers, and theme naming — each a separate, schema-constrained call.
Vector search
Every note, chunk, tag, and entity gets an embedding. Postgres + pgvector (HNSW-indexed) power semantic search and similarity-based canonicalization.
Entity & tag canonicalization
New tags and entities are matched against existing ones by embedding distance before creating a duplicate — so your vocabulary stays clean as it grows.
Next.js + Postgres
Server-rendered throughout, backed by hosted Postgres (Supabase) with Prisma. No separate vector database to keep in sync.
Enforced by the database, not just the app.
Passwords are hashed, never stored in plain text. Reads are enforced with Row-Level Security at the database itself — not just trusted to application code — so Postgres refuses to return another account's rows even if the app had a bug. Export your full knowledge base as JSON or CSV whenever you want.
See the full security model →Questions worth answering up front
How is a hidden tag different from a regular tag?+
Visible tags are the ones you write yourself — a name, a project, a date, as many or as few as you want. Hidden tags are a second layer Taggard builds automatically: latent frames, tone, and methodology, each with a confidence score — signals no one would bother tagging by hand, and ordinary search can't see. They're what powers cross-reference detection and theme clustering.
Does Taggard replace my note app?+
No. Taggard sits underneath whatever you already use to capture notes — drop in a document or paste text, and it builds the second memory layer on top, without asking you to write anywhere new.
Can I still browse by my own tags only?+
Yes. The tags you write are always there for ordinary browsing and filtering — the hidden layer runs underneath without changing how you search when you don't need it.
What happens to my data?+
It's isolated at the database level today via Row-Level Security, with a roadmap toward on-device embeddings and a fully offline desktop option. Taggard never sells your data and never trains models on it.
Feed it one note. Watch the second layer build itself.
Each note you add gets tagged, resolved against your existing entities, and checked for connections to everything else you've written — automatically. Start with anything: a note, a PDF, a URL.