Everything you already have, connected.
Drives, mailboxes, chat platforms, CRMs, issue trackers, SQL databases: every kind of source, one read-only credential each. Connect once; the knowledge base stays current without anyone tending it.
Google Drive
INDEXED 14:32
Microsoft Teams
DIFF · 3 CHANGED
Notion
FETCHING
Salesforce
QUEUED
SQL database
RECORDS · 214
Jira
INDEXED 14:17
Dropbox
PRUNED · −2
Fresh on its own schedule.
Each source is swept incrementally: the source is listed, checksums are compared, and only changed items are fetched and re-indexed. Removals propagate; failures don't, a broken credential is never read as "everything was deleted".
Every 15 minutes by default; manual sync any time
Provider checksums and revisions, unchanged items are not re-read
PDF, Word, Excel, PowerPoint, HTML, Markdown, CSV, plain text
Connect, choose access groups, sync, three clicks per source
Your business records, readable to the assistant.
SQL databases, CRMs and ERPs are read read-only and rendered into records the assistant can reason over: each customer, deal or invoice becomes a dated record carrying its relationships, its line items and exact figures, totals are computed at sync, never estimated by a model.
Column meaning is inferred from types and values, not names, a cryptic legacy schema or a Greek registry works as well as a clean English one. Issues and tickets arrive as full documents with their comment threads.
Read-only sessions with statement timeouts, the schema is never touched
Computed exactly at sync: "invoices: 2 · total €7,000 · latest 14 May"
Records re-render on every sweep; a changed row is a changed record
Sensitive columns, passwords, card numbers, IBANs, salaries, are never ingested
- Standing
- Active since 2023 · renewal in legal review
- Invoices
- 2 open · total €7,000 · latest 14 May
- Deals
- Renewal Q3, 12% pricing exception, approved March
- Tickets
- #4127 invoice import · #4131 onboarding, answered
Connecting company data to AI, answered
What data sources can a company knowledge engine connect to?
Everything that holds company knowledge: drives, mailboxes, chat platforms, CRMs and ERPs, issue trackers and SQL databases. Each connection takes one read-only credential and three clicks, and the live demo shows the catalog connected, source by source.
Do we have to prepare or clean our documents first?
No. Connect sources as they are. The assistant reads the formats companies actually use, Office documents, PDFs, plain text and the rest, and keeps up with changes on its own.
How does the knowledge base stay up to date?
On its own: every source re-checks itself every 15 minutes, picks up what changed and lets removals leave too. A manual sync is one click when you cannot wait. The exact sweep mechanics are in the docs if you want them.
Is connecting our systems safe?
Yes. Every connector is read-only, the assistant can never modify, move or delete anything in your systems, and a credential you enter is used for syncing and never displayed again. The security page carries the full record for your IT lead.
Can AI read our SQL database or CRM without risk?
Yes, that is a first-class case, not a hack. Connections stay read-only, sensitive fields like card numbers and salaries are excluded by rule, and your rows become dated records with exact figures, computed, never estimated by a model. The docs walk through databases and CRMs one by one.
What happens when a connection breaks?
Nothing dramatic: a failed sweep is never read as "everything was deleted". The source keeps its last good state, the console shows the failure, and the next successful sweep carries on incrementally.
Watch a corpus stay fresh.
The demo is a real deployment with its sources connected. Open it and look at the sync log yourself.