Getting started

1. Create your account and workspace

Start free and name your workspace - most teams use their organization or team name. Everything you add is isolated to your account, and nothing you upload is ever used to train AI models.

2. Add your first source

There are two ways to feed the map, and you can use both:

• Connect a tool. Go to Sources and connect Google Docs, Jira, Notion, Confluence, GitHub, or Dropbox. After a one-time consent screen, Rationale syncs new and changed artifacts on its own - see Connect your tools.
• Upload directly. Drop in individual files (.txt, .md, .pdf, .docx, or .vtt meeting transcripts) - useful for a quick trial with one meaty document, like an RFC or a postmortem.

A good first artifact is one where a real decision was made: a design doc with alternatives, an architecture RFC, an incident postmortem. The richer the reasoning in the document, the more the map has to work with.

3. Let Rationale read

Each artifact goes through the understanding pipeline: Rationale identifies the decisions inside it, the evidence and assumptions behind them, the alternatives that were considered, and how it all connects to what's already in your map. Processing typically takes a little while per document - you can keep adding sources in the meantime. The details are in How Rationale reads your artifacts.

4. Explore your map

• Decisions - every decision Rationale has reconstructed, with its problem, evidence, tradeoffs, and alternatives.
• Decision Map - the graph view: how decisions inform, depend on, and supersede each other.
• Timeline - the same knowledge in time order, so you can see how a decision evolved.
• Attention - what Rationale thinks deserves a human look: conflicting evidence, stale decisions, and items flagged for review.

Every claim you see traces back to its source artifact, and every decision shows how many distinct documents support it - single source, corroborated, or verified by a person - so you can always judge its weight and check the original.

5. Review what Rationale isn't sure about

The map is built by AI, but AI output is never auto-trusted - you have the last word:

• Held documents. Anything the sensitive-content screening flags is held before an AI provider ever sees it. Review the findings and approve or exclude each document in one click - see sensitive-content screening.
• Attention items. Conflicting evidence, stale decisions, and anything else that deserves a human look queue up in Attention.
• Corrections. On any decision you can verify it, correct it, or delete it. Corrections outrank what re-reading a document says, and deletions are remembered - a future sync never brings back what you removed.