Connected
A knowledge layer, not another file pile
Knemora compiles approved sources into linked Markdown pages you can search, browse, and carry with you.
Open-source. Local-first. Yours.
Your knowledge. Connected.Your AI. Local.
Turn documents and datasets into linked knowledge with an open-source, local-first AI (SI) Second Brain, bounded AI (SI) Agents, a native graph, and cited answers.
Local inference Portable Markdown MIT open source
Local workspace
Sources
3 readyConnected knowledge
Local Cited answer
Aurora’s review requires local processing and encrypted project storage.
Built around your real sourcesPDFTXTMDDOCXCSVXLSX originals preserved
A durable second brain
Private knowledge should become more useful—not less yours.
Inspectable
Answers that lead back to evidence
Citations keep the Wiki path, original filename, and source locator close to every grounded answer.
Local-first
Private inference on your workstation
Ordinary document processing and model inference stay on your trusted local machine—without a mandatory cloud LLM API.
Portable
Readable formats you continue to own
Originals, Markdown knowledge, SQLite metadata, and portable backups keep your work inspectable and movable.
HOW IT WORKS
One source. Four visible handoffs.
A vertical route makes each change of state explicit—from preserved bytes to a cited answer.
- 01
Add sources
PDF, TXT, Markdown, DOCX, CSV, and XLSX.
- 02
Extract & normalize
Preserve originals and create traceable records.
- 03
Compile knowledge
Validate model proposals into a linked Markdown Wiki.
- 04
Explore & ask
Search, follow relationships, and get cited local answers.
Product walkthrough
See knowledge take shape.
One fictional project, four real interactions. Follow source files as Knemora preserves them, compiles a linked Wiki, and returns a locally generated answer with evidence.
Interactive product preview — sample data
Immutable originals
Start with the sources you already trust.
This packaged preview loads sample files only. It never asks website visitors to upload private documents.
PDFTXTMDDOCXCSVXLSX
PDFSample
Project_Aurora_Brief.pdf
1.8 MB · original available
TXTSample
Security_Review.txt
24 KB · original available
CSVSample
Equipment_Budget.csv
12 KB · original available
Native Second Brain
Follow the links. Read the knowledge.
Knemora combines a native relationship graph with readable Wiki pages. The graph helps you move through explicit links; the page keeps claims and citations understandable.
Aurora workspace
6 pages · 8 explicit links
Global graphSample data
Local AI (SI), clear boundaries
Private by default. Public when you choose.
Documents, the Wiki, operational metadata, and ordinary inference stay on your workstation. Optional public search is a separate, explicit path—not a hidden background service.
Prepared private operation can work offline
Initial setup still needs internet access for dependencies, runtime components, and model acquisition.
Ingestion builds knowledge—not a new model
The model proposes structured knowledge; application code validates and persists it.
Trusted local workstation
Ordinary private path
Sources + Wiki
Bounded tools
llama.cpp model
Cited private answer
Optional public search
Off by default · launch enabled · explicitly requested per turn
Ownership & portability
Built to remain understandable.
Knemora’s durable pieces use formats you can inspect and move. Search and graph indexes remain rebuildable projections—not proprietary sources of truth.
Markdown Wiki
Human-readable pages, claims, links, and provenance.
Preserved originals
Byte-identical source material with traceable identity.
SQLite metadata
Local operational state without a database server.
Portable backups
Knowledge and durable data, excluding model weights and secrets.
Open source · MIT
Start a private workspace on your Mac.
Automatic setup is currently verified for macOS on Apple Silicon with Metal. Bring Node.js 24+, npm, and at least 7 GB free.
No additional system dependencies required from you
Terminal
# after obtaining the application checkout
$ ./knemora setup
# start private local services
$ ./knemora start
localhost · ready
Built for macOS today. More systems next.
Available now
macOS
Apple Silicon · Metal
Coming soon
Windows
Automatic setup
Coming soon
Linux
Automatic setup
Current setup instructions are maintained in the Knemora repository.
FAQ
Clear answers, before you begin.
What is Knemora?
Knemora is an open-source, local-first Second Brain for organizations and people working with private documents and datasets. It turns source material into durable, searchable, interconnected knowledge.
Where does AI (SI) inference happen?
Ordinary private inference runs locally through llama.cpp on your workstation. Knemora does not require a hosted LLM API or inference subscription for normal private use.
Does setup need an internet connection?
Initial setup needs internet access to acquire dependencies, managed runtime components, and the selected model. Once prepared, private operation can work offline.
How does optional public search work?
Public search is off by default. The operator must enable it at launch and the user must explicitly request it in the current turn. Only a bounded public query leaves the machine; public results stay separate from private Wiki evidence.
Which source formats are supported?
Knemora currently accepts PDF, TXT, Markdown, DOCX, CSV, and XLSX sources.
Can I take my knowledge with me?
Yes. The knowledge layer is a human-readable Markdown Wiki, original source bytes are preserved, operational metadata uses SQLite, and portable backups cover the durable knowledge and data needed to move a workspace.
Which platforms are supported today?
Automatic zero-configuration setup is currently verified for macOS on Apple Silicon with Metal. It requires user-installed Node.js 24 or newer with npm. Windows and Linux automatic setup remain future work.
Make knowledge durable
Your sources already hold the context. Connect it.
Knemora turns private documents and datasets into a local Second Brain you can inspect, explore, and keep.
Review local setup