What you'll do#
Register access to a system (for example, your ERP's database) and get it ready for a pipeline. The connection stores the address and the password (encrypted); the pipeline is what brings in the data using that access.
Registration has four steps, always in this order:
| # | Step | What you do in it |
|---|---|---|
| 1 | System | Choose where your data lives (Postgres, MySQL, SQL Server, Oracle, and other databases). |
| 2 | Address and password | Fill in the server address, database name, user, and password. The connection name comes suggested. |
| 3 | Test | Click Test now and read the result in one sentence. |
| 4 | Done | Review and save. The password isn't shown. |
The pipeline points to the connection, it doesn't copy the password
Changing the password in one place applies to every pipeline that uses the connection.
Before you start#
- Log in to the right account and check the company/workspace shown in the app.
- You need to be the workspace owner: ETL/ELT → Connections is an owner area.
- Choose a source offered in the
connector catalog and read its page. A
connector marked
Limited accessmay require enablement; an item markedPlannedis not a usable connection yet. - Ask whoever is responsible for the source for access suited to the task. To read data, don't ask for system administration permissions.
- Number of connections per plan: Starter 3 · Growth 10 · Business 30 · Scale 100 · Enterprise custom. See Plans.
- Registering a connection doesn't start any ingestion and doesn't generate cloud cost.
If you're only going to upload a file, use the import flow — see import a file. A local file doesn't require making up a password or creating a vault.
Where it starts#
In ETL/ELT → Connections. Before creating one, look at the list: if there's already a connection to the same system, use it. Each connection shows a Status — Working, Has a problem, Never tested, or Test again.
Use New connection to open the system choice.
Step 1 — System#
Choose where your data lives. Each card says, in one line, what we'll ask for in the next step. If you arrived from a catalog card, this step is already done.
Is your data in a spreadsheet? You don't need a connection: create a pipeline and upload the file directly.
Step 2 — Address and password#
Fill in what your IT team gave you. Each field has a help line with an example:
- Server address — name or IP of the database server.
- Port — the default is pre-filled; only change it if IT says it's different.
- Database name, User (preferably a read-only one, created for BI), and Password.
The connection name comes suggested (system + database name); change it if you like. Never put a password or personal data in the name.
Advanced options (collapsed) holds what usually only IT touches: schema, connection security (SSL), how we reach the database (directly over the internet, SSH tunnel, Cloud SQL, VPN). It comes with the secure default.
The password is stored encrypted and never shows on screen again — not even for you. To change it later, use Change the password.
If something goes wrong here: the field with the problem turns red, with the explanation below it and a summary at the top.
Step 3 — Test#
Click Test now. We try to log in to the system with the data provided and list the tables — nothing is changed in your system.
The result is one sentence: Connection working (with how many tables we found) or We couldn't get in, with the reason and what to do — for example, "The server didn't respond in time… ask your IT team to allow access". The per-stage detail (network, user and password, read) is under See test details (for IT).
The test is optional: Continue without testing (or Continue anyway, after a failure) saves the connection as Never tested.
Step 4 — Done#
Review the data and save. The password isn't shown, only whether it was provided.
Use Save connection. If you came from building a pipeline, the button is Save and use in pipeline and you go back there with the connection already selected. Cancel saves nothing.
Saving doesn't bring in data: the pipeline that uses the connection does that.
The four states behind the status (for IT)#
The list shows ONE status. The connection's detail page shows the four facts it comes from — and a connection can be "Active" and never have brought in anything:
| State | Possible values |
|---|---|
| Registration | Draft · Active · Archived |
| Validation | Validated ✓ · Failed the test · Testing… · Validation outdated · Not tested |
| Credential | Credential configured · Credential expired · Credential revoked · Credential inaccessible · No credential |
| Last ingestion | Last ingestion OK · Last ingestion failed · Ingestion running · Never ingested · No pipeline |
Expected result#
You have an identifiable, authorized connection; the credential is encrypted and referenced; the test status matches what was actually verified; the ingestion state says Never ingested or No pipeline — because saving doesn't load any data.
If something goes wrong#
| Symptom | What to check | Next action |
|---|---|---|
| Network failed | Address, port, network mode, allowlist | Allow the application server and the executor; validate with whoever manages the source |
| Authentication failed | Method, validity, and user chosen | Replace or reauthorize the credential as the connector requires |
| Authenticates, but read failed | Permissions on the specific schema/table | Ask only for the access you need (read-only) |
| Read "Not tested" | Whether the connector supports that check | Treat it as not verified, not as approved |
| The next step doesn't open | The error summary at the top of the current step | Fix the field in red |
| Connector unavailable | Public status on the connector page | See connector catalog |
| I don't see Connections in the menu | Your role in the workspace | Connections is an owner area; ask the owner for access |
| I hit the connection limit | Plan limit | Archive an unused connection or see Plans |
Limits and costs#
- Registering, testing, and saving: R$ 0 on the platform. Some external connectors charge per API call — check the connector page.
- The number of connections is limited by the plan.
- We don't publish registration or test times: they depend on the source and the network.
Back to the app#
- Register now: Connections → New connection
- See the connections already registered: Data connections
Next task#
Create the pipeline that will use this connection: First ingestion. On the connection's detail page, the shortcut is Use in a new pipeline.
Last reviewed: 2026-10-03. The screenshots on this page come from the delivered
version, in a demo environment with fictional data (the project seed's synthetic
workspace), 1366×768 viewport. The provenance of each image
— file, SHA-256, route, commit, role, and privacy review — is in the
screenshot manifest e2e/baseline/galeria-v3.json. A screenshot from a local
environment documents the interface; it doesn't prove execution in the cloud. No customer data,
names, or credentials appear in this documentation.