> ## Documentation Index
> Fetch the complete documentation index at: https://docs.inviolet.ai/llms.txt
> Use this file to discover all available pages before exploring further.

# Connect your first data source

> Salesforce, Snowflake, or Postgres — the three most common starts.

Once the gateway is installed and observing tool calls, the next step is
connecting the data sources your agents actually query. This unlocks the
column-level enforcement layer and the Vault credential brokering path.

<Note>
  Phase 1 (anonymous observation) does NOT require a data source connection.
  The gateway sees tool calls regardless. Connecting a source is required to
  enforce policy at the database layer (Phase 6 — DB proxy).
</Note>

## Pick a starting source

Most teams start with whichever source already has the most LLM tool-call
volume:

* **Salesforce** — sales / customer support agents
* **Snowflake** — analytics / data-team agents
* **Postgres** — transactional / product agents

The connection is the same shape for all three.

## Connect via the dashboard

1. Sign in at [app.inviolet.ai](https://app.inviolet.ai)
2. Settings → **Data Sources** → **Add new**
3. Pick the connector type
4. Provide credentials (read-only is fine for the observation phase)
5. Inviolet performs a test connection + lists discovered tables
6. Confirm — the source is live

## Watching the heatmap fill

Once the source is connected, the **Query Pattern Map** updates as your
agents make tool calls. After 50–100 calls you'll see which tables and
columns get touched by which intentions.

## Promoting to enforcement

When the heatmap is stable for 2+ weeks, you can declare an intention against
this source — see
[Define your first intention](/guides/define-your-first-intention).

## Read next

* **[Define your first intention](/guides/define-your-first-intention)**
* **[Generalize from observation](/guides/generalize-from-observation)**