You can connect external data sources to give MAIA direct access to your organization's documents and records. This is entirely optional — MAIA works without connected sources, using only what you describe and upload.
Supported Sources
MAIA supports connecting to:
- Google Drive — Documents, spreadsheets, presentations
- Salesforce — CRM records, accounts, opportunities
- Slack — Channel messages, relevant discussions
- HubSpot — Marketing and sales data
- Jira — Project tracking, tickets
- Microsoft (OneDrive, SharePoint, Outlook, Teams) — Documents, email, communications
- Dropbox — Files and folders
- Box — Enterprise content
- Asana — Project and task data
How Much You Can Upload
A rehearsal analyzes up to 60,000 characters of uploaded documents in total — roughly 25–30 pages of text. MAIA tells you at upload if a file did not fit.
A file is included whole or set aside whole; MAIA never uses part of a document without telling you. If a file is too large, split it into smaller files and upload the section that matters most to this decision — a single chapter, the financial tables, the relevant appendix. Uploading the same oversized file again will give the same result.
Brick-by-Brick Knowledge Architecture exports are the exception: they come in through their own channel, are read structurally rather than as raw text, and do not count against this limit.
How Authorization Works
When you connect a source, you're taken through a standard OAuth flow — the same "Grant access" screen you see when connecting any third-party app. You grant MAIA read-only permission. MAIA never writes to, modifies, or deletes anything in your connected sources.
You can revoke access at any time from your account settings or from the source's own app permissions page.
What MAIA Pulls
MAIA doesn't download everything in your account. It uses your decision description to construct targeted search queries and fetches only the documents and records relevant to your rehearsal. If you're rehearsing a market expansion decision, MAIA looks for market analysis documents, competitive intelligence, financial projections — not your team's lunch plans.
Data Handling
- All fetched data is cached for a maximum of 24 hours, then permanently deleted
- Fetched content is screened for sensitive data (PII, PHI, credentials, classified markings) before it enters the pipeline. Flagged content is excluded automatically.
- Connected source data follows the same Zero Data Retention policy as everything else in MAIA — processed in memory, never written to permanent storage
- Your OAuth tokens are stored securely and are only used during active rehearsal sessions
For more on how MAIA protects connected source data, see Zero Data Retention.