diff --git a/docs.json b/docs.json
index b1efc4c..627b353 100644
--- a/docs.json
+++ b/docs.json
@@ -294,6 +294,7 @@
"features/data-grid",
"features/query-results",
"features/filtering",
+ "features/search-database",
"features/change-tracking",
"features/json-viewer",
"features/data-files"
diff --git a/features/search-database.mdx b/features/search-database.mdx
new file mode 100644
index 0000000..424564a
--- /dev/null
+++ b/features/search-database.mdx
@@ -0,0 +1,85 @@
+---
+title: Search database
+description: Find every table and column that holds a value, such as an email address, an order number or a UUID, and open the matching rows
+---
+
+Type the value once and every table is checked for it, one `SELECT` per table. Each result names the
+columns that hold the value and opens the table already filtered to those rows.
+
+
+ 
+ 
+
+
+## Running a search
+
+
+
+ Choose **Database > Search Database…** to search the database being browsed. In the sidebar,
+ right-click a database row and choose **Search Database…**, or a schema row and choose
+ **Search Schema…**.
+
+
+ Type it in **Find**, pick a **Match** mode, and select **Match Case** to tell `Ada` from `ada`.
+ Where the column's collation decides case, **Match Case** is dimmed; see
+ [Case sensitivity](/features/filtering#case-sensitivity).
+
+
+ Every table in the database is selected, or every table in the schema you right-clicked. A
+ schema's checkbox selects or deselects all of its tables. **Filter Tables** narrows the list,
+ and **Select All** and **Deselect All** act on what it shows.
+
+
+ A progress bar counts the tables done out of the total. **Stop** ends the search after the
+ table being read, and keeps what was found so far.
+
+
+
+**Contains** is the mode to start with. Use **Matches Exactly** for an identifier such as an order
+number, so a longer value that contains it does not match.
+
+## Which columns are searched
+
+| Column type | Searched |
+|---|---|
+| Text | Always, with the chosen **Match** mode |
+| Enum, set | With **Contains** and **Starts With**; with **Matches Exactly** when the value is one of the labels |
+| JSON | With **Contains** and **Starts With**, matched against the document's text |
+| Integer, decimal | When the value is a number, as an exact comparison |
+| Date | When the value is a date such as `2024-02-29`, as an exact comparison |
+| Timestamp, datetime | When the value is a date and time such as `2024-02-29 13:45:10`, as an exact comparison |
+| UUID | When the value is a UUID, as an exact comparison |
+| Binary | Only with **Include binary columns** selected, and never with **Matches Exactly** |
+| Boolean, time, array, spatial | Never |
+
+With **Matches Exactly**, large text types such as `CLOB` and `NTEXT`, and types stored as text such
+as `inet`, are skipped. A table with no column to search is left out of the results.
+
+Enum, set, JSON and binary columns are searched on the MySQL, PostgreSQL, SQLite, SQL Server,
+ClickHouse and DuckDB families, and skipped on Oracle. On Trino, Snowflake, BigQuery and the other
+engines outside those families, only text and number columns are searched.
+
+## Reading the results
+
+Each table that holds the value shows the matching columns and how many rows match. Counting stops at
+1,000 rows per table, shown as **1,000+ rows**. A table that could not be searched, for lack of
+permission or because the query timed out, is listed with the server's error. Each query runs under
+the connection's query timeout.
+
+Double-click a table, or select it and click **Open**, to open it in a new tab. The filter bar holds
+one condition per matching column, joined with **Match any**, so the tab shows exactly the rows the
+search counted. When the count reads **1,000+ rows**, the tab gets a condition for every searched
+column instead.
+
+Choose **Search Database…** again on the same database to come back to the results. **Edit Search**
+returns to the form with the value and the tables you chose.
+
+
+Search runs on databases queried in SQL. It is not offered on Cassandra, ScyllaDB or DynamoDB, nor
+on MongoDB, Redis and the other engines that do not use SQL.
+
+
+## Related
+
+- [Filtering](/features/filtering): the filter bar a result opens with
+- [Open Quickly](/features/open-quickly): jump to a table, view or routine by name
diff --git a/images/search-database-dark.png b/images/search-database-dark.png
new file mode 100644
index 0000000..70ba4c3
Binary files /dev/null and b/images/search-database-dark.png differ
diff --git a/images/search-database.png b/images/search-database.png
new file mode 100644
index 0000000..ef9d9e1
Binary files /dev/null and b/images/search-database.png differ