Databases & CRM
Set up MongoDB
Connect a MongoDB database with a read-only user. Mongo has no schema catalogue, so recent documents of each collection are sampled to learn its fields; every collection becomes a searchable entry, and collections that look like registries (they carry a name-like field) also become one record per document. Read-only — only finds are ever issued.
Before you start
- A database user with the read role on the database you want read (Atlas: Database Access → Add new database user → read) — never a role that can write.
- The connection string for that user (Atlas: Connect → Drivers → the mongodb+srv://… string), with the password filled in.
- Documents → MongoDB: paste the Connection string.
- Set Database only if the string doesn't already name the database at the end of the path.
- Only these collections narrows to specific collections; Skip these collections leaves particular ones out instead — blank on either reads every collection that isn't a system one.
- Choose access groups, Test, then Sync.
What comes in
- Every collection as a searchable entry — its fields (sampled from the 200 newest documents), their types, and a few real sample values.
- Collections that carry a name-like field also become one record per document — up to the newest 2,000, ordered by _id.
- Nested documents are flattened one level (a.b); anything deeper is summarised as a count, never spelled out.
| Database | Which database on the deployment to read — blank uses the one named in the connection string. |
|---|---|
| Only these collections | Allow-list, one collection per line; blank reads the whole database. |
| Skip these collections | The opposite — read every collection except these. |
Access
Visible only to the access groups chosen on the source — MongoDB has no per-document viewer list to mirror, so that source-wide list covers everything a collection brings in.
Sensitive fields — passwords, tokens, card numbers and the like — are never ingested, whatever collection they're in. A collection past 2,000 documents syncs only its newest 2,000 as records — its schema-level entry stays searchable regardless — and a database past 200 collections indexes only the first 200, alphabetically; both are named when they happen.
Last updated 20 Sep 2026