<!-- Source: https://docs.snoutdata.com/databases/mongodb -->

# MongoDB

MongoDB is a **document** database. SnoutData connects to it like any other database: its
collections show up in the explorer, you can query them with SQL or a native aggregation
pipeline, edit documents in the results grid, and manage collections and indexes.

## Connect

1. Open the **connections sidebar** and choose **New connection**.
2. Pick the **MongoDB** driver.
3. Provide either a full **connection string** (`mongodb://...` or `mongodb+srv://...`) or a
   host, port, and authentication database with a username and password.
4. Click **Test**, then **Save**.

Flag the connection as **read-only** or **production** if you want to block writes. See
[Security](https://docs.snoutdata.com/connections/security).

## Browse collections

Expand the connection to see its **collections**, listed like tables. SnoutData infers a
collection's fields by **sampling documents**, so each collection shows its top-level field
names and types. Views appear too. Because documents in a collection can differ, the field list
is a best-effort picture of the shape, not a fixed schema.

## Query with SQL

Write SQL against a collection and SnoutData transcribes it to an aggregation pipeline. The
common shapes are supported:

```sql
SELECT name, email
FROM users
WHERE country = 'US'
ORDER BY created_at DESC
LIMIT 50;
```

`WHERE` (including `=`, `!=`, comparisons, `IN`, `IS NULL`, and `LIKE`), projections, `ORDER BY`,
`LIMIT` / `OFFSET`, and `GROUP BY` with `COUNT` / `SUM` / `AVG` / `MIN` / `MAX` all translate to
the matching pipeline stages. Anything the offline compiler cannot represent can be translated by
the AI assistant after a confirmation prompt (see [querying beyond SQL](https://docs.snoutdata.com/databases/overview#when-sql-is-too-complex-to-translate-offline)).

:::tip See the pipeline for your SQL
Select a SQL statement, right-click, and choose **Get Pipeline Translation** to insert the
equivalent aggregation pipeline as a comment. It is a quick way to learn the pipeline for a query
you already know, or to start a native pipeline from it.
:::

## Write a native pipeline

For the full power of MongoDB, write the aggregation pipeline directly. On a MongoDB connection
the editor toolbar shows a **SQL / Pipeline** switch and a **collection** picker. Flip to
**Pipeline** and write a JSON array of stages:

```json
[
  { "$match": { "country": "US" } },
  { "$group": { "_id": "$plan", "users": { "$sum": 1 } } },
  { "$sort": { "users": -1 } }
]
```

The pipeline editor has schema-aware completion for stage operators, the collection's field
names, and collection names (for `$lookup`). Run it like any query; results land in the same
grid. Switching between SQL and Pipeline keeps both buffers, so you can move back and forth.

## Edit documents

On a non-production, writable connection you can edit data straight from the [results grid](https://docs.snoutdata.com/editor/results-grid):

- **Edit a cell** to update that field on the document (`updateOne` keyed on `_id`).
- **Insert** a new document, or **delete** one.

Edited values are coerced to the field's inferred type (numbers, booleans, and so on). Editing
requires each row to have an `_id`. A few limits to know:

- Date, nested-object, and array cells edit as raw strings (an explicit "set to current time"
  produces a real date).
- A save with several changes is applied one operation at a time, not as a single transaction,
  so a failure partway through can leave some changes applied. SnoutData notes this before you
  confirm.

## Manage collections and indexes

A MongoDB connection's **Properties** tab shows the collection's inferred fields and its indexes,
where you can:

- **Create** or **drop** a collection (the per-database **+** in the sidebar is **New
  collection**).
- **Create** or **drop** a secondary index. On a new or empty collection you can type the field
  names to index by hand.

Each change shows its native command and asks for confirmation before it runs.

## Ask the assistant

On a MongoDB connection the [AI assistant](https://docs.snoutdata.com/ai-assistant/overview) proposes a native
**aggregation pipeline** instead of SQL, grounded in the connection's collections and fields.
Insert it into a pipeline tab and run it. In [agentic](https://docs.snoutdata.com/ai-assistant/agentic) mode the
assistant can run a read-only pipeline for you and read the results.