Connect your tools

How connecting works

• In the app, open Sources and choose Add source. Pick a provider and select Connect.
• You'll be sent to the provider's own consent screen (Google, Atlassian, Notion, GitHub, or Dropbox) to approve read-only access. Rationale never writes to, edits, or deletes anything in your tools.
• After you approve, you land back in Rationale and choose what to sync: pick specific projects, spaces, folders, or repositories (with a live count of the documents your choice covers), write an advanced filter, or explicitly choose everything. Nothing syncs until you decide - a narrow, curated scope beats the whole site.
• Once scoped, the source starts syncing and Rationale checks for new and changed artifacts automatically from then on. You can edit the scope at any time from the same screen.
• Access tokens are stored encrypted, and you can disconnect a source at any time from the same screen - its content stops flowing immediately.

Google Docs

Rationale asks for read-only access to your Google Drive - Google's consent screen describes this as seeing your Drive files. After connecting, pick the Drive folder to sync from the scope step. Rationale syncs only your Google Docs, keeps them current as they change, and never edits anything. During early access the Google connector runs under a test registration, which means accounts need to be allowlisted first - contact us and we'll enable yours.

Jira

Connecting Jira goes through Atlassian's consent screen, where you pick the site to connect. Back in Rationale you then choose which projects to sync (or write a JQL filter). Rationale reads issues and their discussion - often the richest record of why something was decided. If Atlassian shows "Access denied - requires a site", the account you consented with has no Jira site attached; sign in with an account that belongs to your Jira site and try again.

Confluence

Confluence uses the same Atlassian consent flow as Jira - you approve each product once, with the same account, then choose which spaces to sync (or write a CQL filter). Rationale reads pages and spaces, which is where most teams keep design docs, RFCs, and runbooks.

Notion

Notion's consent screen asks you to choose which pages to share with Rationale - only the pages you select (and their subpages) are synced. You can change the shared set later from Notion's own connection settings.

GitHub

Connecting GitHub installs the Rationale GitHub App on the repositories you choose - you pick exactly which repos during the install, and can change the set later from GitHub's own settings. The scope step in Rationale can narrow further: a subset of the installed repos, or path patterns like docs/**. Rationale reads only documentation files (Markdown and MDX): ADRs, design docs, and docs folders - including sites built with tools like Docusaurus, MkDocs, or Mintlify, whose content lives in your repo. Code is never ingested, and access is read-only.

Optionally, GitHub can also capture merged pull request discussions - the PR description, review threads, and approval rationale, which is where many engineering decisions are actually made. This is off by default: turn it on per connector from the Rules tab in the scope settings. Only merged PRs are captured (a merge is the signal a decision was actually taken), bot comments and review nitpicks are filtered out, and the backfill reaches 180 days by default. Enabling it may require approving updated GitHub App permissions (Pull requests and Issues, both read-only); until you approve, document syncing continues unaffected and the connector shows what is pending.

Dropbox

Dropbox's consent screen grants Rationale read-only access to your files. Rationale syncs document files only (PDF, Word, Markdown, and text). After connecting, pick the folder to sync from the scope step - the connector reads that folder and nothing else (or choose everything explicitly). Files you delete in Dropbox stop syncing; what was already extracted stays in your Decision Map until you remove that artifact in Rationale.

Sensitive-content screening

Every document - uploaded or synced - is screened locally before anything is embedded, indexed, or sent to an AI provider. Deterministic pattern detectors look for structured identifiers (validated card numbers, SSNs, IBANs), legal privilege markers, bulk contact rosters, and combinations of compensation, classification, and board-material vocabulary. A flagged document is held for your review with the findings shown (masked, never the raw values): approve it in one click, or exclude it - excluded content is deleted and stays excluded across future syncs. You can also set per-connector deny rules (paths, labels, title patterns) so matching documents are never stored at all; connectors report how many were excluded by your rules. To be precise about the limits: this is pattern matching, not semantic data-loss prevention - choosing what to sync remains the primary control.

Privacy notes

• All connector access is read-only.
• Your content is never used to train AI models.
• Held documents are never sent to an AI provider, and nothing is auto-deleted without your say-so.
• Disconnecting a source stops its syncing immediately, and you can request erasure of a person's data - see our privacy policy.