Cornerstone · Everyday use
Load your documents
Everything Cornerstone knows comes from documents you load. There are two ways in, both on the Sources page.
Upload a file
- Open Sources.
- Choose the project the document belongs to, then Upload a document and pick the file.
- A line under the button confirms it: stored, or already stored if the same file was loaded before.
Text files (Markdown, plain text, CSV, JSON, YAML) and Word, PowerPoint, Excel and PDF files become searchable as soon as they arrive. Other files, and scanned PDFs with no text in them, are kept as the original so you can download them, but they cannot answer questions.
If a document with the same name has changed since you last loaded it, the upload is refused with the reason rather than overwriting someone else's edit.
Connect a Git repository
Choose Connect a source, or open Connections. Cornerstone uses a GitHub App installed on your organisation, so there is no token to paste and you can revoke it from GitHub at any time. A connected repository re-syncs on every push, and hourly as a backstop.
From searchable to answerable
A loaded document can be searched straight away. To answer questions with confidence, Cornerstone needs it turned into short facts: rules, decisions, processes, definitions and references. This step is called distilling, and your AI does it: ask it to distil the document, and it proposes the facts. See Review proposed facts.
Quality shows how much of your workspace has been distilled and which documents are still waiting.
What to keep out
Do not load personal data you have no lawful basis to process, or special-category data such as health records, unless your organisation has a valid condition for it.
Updated 14 November 2026