diff --git a/config/_default/menus/main.en.yaml b/config/_default/menus/main.en.yaml
index b50ecdb5347..310d0e547a8 100644
--- a/config/_default/menus/main.en.yaml
+++ b/config/_default/menus/main.en.yaml
@@ -4635,6 +4635,11 @@ menu:
       parent: dbm
       identifier: dbm_setup_architecture
       weight: 2
+    - name: Oracle Configurations
+      url: database_monitoring/architecture/oracle/
+      parent: dbm_setup_architecture
+      identifier: dbm_architecture_oracle
+      weight: 4
     - name: Setting Up Postgres
       url: database_monitoring/setup_postgres/
       parent: dbm
diff --git a/content/en/database_monitoring/architecture/oracle.md b/content/en/database_monitoring/architecture/oracle.md
new file mode 100644
index 00000000000..ccc0550310f
--- /dev/null
+++ b/content/en/database_monitoring/architecture/oracle.md
@@ -0,0 +1,175 @@
+---
+title: Oracle Configurations
+description: Configure Oracle instance scoping for Database Monitoring using PDB-level, CDB listener, and CDB$ROOT approaches.
+further_reading:
+- link: "/database_monitoring/architecture/"
+  tag: "Documentation"
+  text: "DBM Setup Architectures"
+- link: "/database_monitoring/setup_oracle/"
+  tag: "Documentation"
+  text: "Setting up Oracle"
+---
+
+## Overview
+
+Each check instance in the Datadog Agent configuration creates one Database Monitoring (DBM) instance. The [database_identifier](/database_monitoring/guide/database_identifier/) template determines what the Agent considers a distinct DBM instance.
+
+The key configuration lever is `service_name`, which determines whether you connect to `CDB$ROOT` or a specific PDB.
+
+Available template variables: `$resolved_hostname`, `$server`, `$port`, `$service_name`, `$cdb_name` (from `v$database`), plus any key from tags.
+
+Common users (C## prefix, created with `CONTAINER=ALL`) are required for multitenant monitoring.
+
+For non-multitenant Oracle (pre-12c or non-CDB), use host-level scoping with `$resolved_hostname`.
+
+**Scenario**: `oracle-prod.example.com` with two Oracle CDBs:
+
+| CDB | Port | PDBs |
+|---|---|---|
+| ORCL | 1521 | CRMDB, ERPDB |
+| FINDB | 1522 | ACCOUNTS |
+
+## Option 1: PDB level (recommended)
+
+<div class="alert alert-info">Use PDB-level scoping (<code>$cdb_name-$service_name</code>). Connecting through each PDB's service name gives isolated DBM data scoped to that PDB's session activity. Add a <code>CDB$ROOT</code> entry alongside PDB entries if you also need server-level Oracle metrics visible only from <code>CDB$ROOT</code>.</div>
+
+One entry per PDB using the PDB-specific service name. Each connection scopes directly into that PDB. Metrics, query samples, and explain plans reflect only that PDB's activity.
+
+Based on the example scenario above, this counts as three DBM instances.
+
+```yaml
+instances:
+  - server: oracle-prod.example.com:1521
+    service_name: CRMDB
+    username: c##datadog
+    password: "<PASSWORD>"
+    dbm: true
+    database_identifier:
+      template: "$cdb_name-$service_name"  # → "orcl-CRMDB"
+
+  - server: oracle-prod.example.com:1521
+    service_name: ERPDB
+    username: c##datadog
+    password: "<PASSWORD>"
+    dbm: true
+    database_identifier:
+      template: "$cdb_name-$service_name"  # → "orcl-ERPDB"
+
+  - server: oracle-prod.example.com:1522
+    service_name: ACCOUNTS
+    username: c##datadog
+    password: "<PASSWORD>"
+    dbm: true
+    database_identifier:
+      template: "$cdb_name-$service_name"  # → "findb-ACCOUNTS"
+```
+
+The Agent automatically adds a `pdb:<name>` tag when it detects a PDB connection.
+
+## Option 2: CDB listener level
+
+<div class="alert alert-info">Use CDB listener level when you need to monitor all PDBs through a single connection per CDB. All PDBs are visible through <code>CDB_*</code> views from <code>CDB$ROOT</code>.</div>
+
+One entry per CDB listener (port). Both CDBs are covered with distinct identifiers.
+
+Based on the example scenario above, this counts as two DBM instances.
+
+```yaml
+instances:
+  - server: oracle-prod.example.com:1521
+    service_name: ORCL
+    username: c##datadog
+    password: "<PASSWORD>"
+    dbm: true
+    database_identifier:
+      template: "$resolved_hostname:$port"  # → "oracle-prod.example.com:1521"
+
+  - server: oracle-prod.example.com:1522
+    service_name: FINDB
+    username: c##datadog
+    password: "<PASSWORD>"
+    dbm: true
+    database_identifier:
+      template: "$resolved_hostname:$port"  # → "oracle-prod.example.com:1522"
+```
+
+## Option 3: CDB$ROOT level
+
+<div class="alert alert-info">Use CDB$ROOT level when logical container names are preferred over network addresses in identifiers. Coverage is the same as Option 2.</div>
+
+Same connections as Option 2, but the identifier uses `$cdb_name` to reflect the logical container name rather than the network address.
+
+Based on the example scenario above, this counts as two DBM instances.
+
+```yaml
+instances:
+  - server: oracle-prod.example.com:1521
+    service_name: ORCL
+    username: c##datadog
+    password: "<PASSWORD>"
+    dbm: true
+    database_identifier:
+      template: "$cdb_name"     # → "orcl"
+
+  - server: oracle-prod.example.com:1522
+    service_name: FINDB
+    username: c##datadog
+    password: "<PASSWORD>"
+    dbm: true
+    database_identifier:
+      template: "$cdb_name"     # → "findb"
+```
+
+`service_name` is the scope selector: CDB service name connects to `CDB$ROOT`, PDB service name connects to that PDB. `$cdb_name` is resolved live from `v$database.NAME`.
+
+## Managed services
+
+### Oracle Autonomous Database
+
+Oracle Autonomous Database (ADB) requires mutual TLS (mTLS) through an Oracle Wallet. Use predefined service names representing workload profiles (`_high`, `_medium`, `_low`, `_tp`, `_tpurgent`).
+
+```yaml
+instances:
+  - server: adb.example.oraclecloud.com:1522
+    service_name: mydb_high
+    username: ADMIN
+    password: "<PASSWORD>"
+    protocol: TCPS
+    wallet: /path/to/wallet_mydb
+    dbm: true
+    database_identifier:
+      template: "$resolved_hostname:$port"
+```
+
+### Oracle Exadata
+
+Exadata is a hardware platform. Apply the scoping approach based on the Oracle architecture deployed on it: CDB/PDB → PDB level; RAC → one entry per node; standalone → host level.
+
+### Oracle RAC (Real Application Clusters)
+
+Connect to each RAC node individually through its local listener. Never use the SCAN address. It may route to a different node between check cycles.
+
+```yaml
+instances:
+  # RAC Node 1
+  - server: rac-node1.example.com:1521
+    service_name: ORCL
+    username: c##datadog
+    password: "<PASSWORD>"
+    dbm: true
+    reported_hostname: rac-node1
+    database_identifier:
+      template: "$cdb_name-node1"
+
+  # RAC Node 2
+  - server: rac-node2.example.com:1521
+    service_name: ORCL
+    username: c##datadog
+    password: "<PASSWORD>"
+    dbm: true
+    reported_hostname: rac-node2
+    database_identifier:
+      template: "$cdb_name-node2"
+```
+
+{{< partial name="whats-next/whats-next.html" >}}