Connect your first source

Goal

One operational data system connected, validated, and syncing. Done means the source appears under Installed with a recent sync time and recognizable sample data.

Image: Integrations Discovery tab with source catalog and search.

Steps

1. Open Integrations in your workspace. It manages sources, destinations, secret access, and external tools — connections live here, not inside individual workflows.
2. Stay on the Discovery tab. It opens first for a reason: pick from the curated catalog instead of configuring blind. Search or filter by the system you use.
3. Choose the source class entry for your system. You will use it as a read point later; workflows reference it, they never embed its secrets.
4. Let setup assist you. After you pick a capability, the assistant helps with configuration: which account to authorize, which databases, schemas, or objects to include, and how wide the first read should be. Start narrow — one database or object, not everything. You can widen later.

Image: AI-assisted source setup showing account, scope selection, and generated configuration preview.

5. Read the configuration preview before anything is saved. Check: the right account, the right scope, and nothing you did not ask for. Access details stay with the integration; workflows only ever see the reference.
6. Validate, then commit. The setup dry-runs the connection — can it reach the system, read the scope, return the expected shape? Only commit when validation passes. A failed validation names the cause (wrong permissions, unreachable host, empty scope); fix that cause, not the whole setup.
7. Confirm under the Installed tab: your source listed, connection healthy, first sync time current. Open the sample and check you recognize the data — right system, right objects, plausible freshness.

Image: Installed tab showing the source healthy with last sync time and sample rows.

When setup fights back

• Validation fails on permissions: the authorizing account lacks read access. Re-authorize with an account that can read the scope — this resolves most first-time failures.
• Empty scope: the connection works but the selected objects hold no data. Widen the scope or pick a different object.
• Stale sync later on: re-check the Installed tab first — connection status and last sync time tell you whether it is credentials, the source, or scheduling.

What good looks like

• One source, healthy, syncing on schedule — not five half-configured ones.
• You can point at the sample data and confirm what each field means.
• Any teammate opening Installed understands what is connected and why.

Next

• Put the source to work: Run your first pipeline.
• Shape what it means: Shape your first ontology.
• If a source breaks, start with Troubleshooting.