Oracle Enterprise Manager and the Oracle Management Agent

Question: which monitoring agent does Oracle AI Database 26ai actually use?

The Oracle Management Agent, shipped with Oracle Enterprise Manager 24ai, currently at Management Agent 24.1.x. You deploy it one of three ways: a local (standalone) Management Agent on the host itself, a Central Agent on the first OMS host, or, new with EM 24ai, a Remote Agent that can monitor a target without an agent installed on that host at all. EM 24ai's Database plug-in covers 26ai targets specifically; if you're monitoring a cloud-hosted 26ai database, such as Amazon RDS for Oracle, only Management Agent 24.1.0.0.v1 is supported there.

Worth being precise about scope here, since the terminology can genuinely mislead: this is a completely different thing from the newer Select AI Agent framework and Private Agent Factory introduced elsewhere in 26ai. Those are in-database, AI/agentic application features, building and running AI agents that work with your data. The Oracle Management Agent covered in this module is host and database monitoring software, part of Enterprise Manager, with nothing to do with AI agents at all. Two unrelated products that happen to share the word "Agent."

The monitoring agent's own history is worth knowing, since older material on this exact topic tends to stop partway through it:

1. Oracle Intelligent Agent (OIA), used in Oracle 8.1.5, 8i, and 9i. SNMP-based communication with the original OEM Console. Deprecated after 9i.
2. Enterprise Manager 10g Agent, introduced with Oracle 10g. Replaced the Intelligent Agent with HTTPS-based communication to the Oracle Management Service (OMS) in OEM 10g Grid Control, the foundation the modern agent architecture still builds on.
3. Oracle Management Agent (OMA), the current line, running from 11g all the way through 26ai today. It works with Enterprise Manager Cloud Control, now at release 24ai, and handles secure bidirectional communication with OMS, host- and database-level metric collection, automatic target discovery, and job scheduling and script execution.

The short version: Intelligent Agent gave way to the 10g Agent, which gave way to the Oracle Management Agent, and that same line is still what's running today, current at Management Agent 24.1.x under Enterprise Manager 24ai.

Since both 19c and 26ai remain active, current releases, it's worth being specific about how they actually compare rather than assuming one has simply replaced the other. Oracle splits database releases into two tracks: Long Term Support releases, which get the longest Premier and Extended Support window, and Innovation releases, which get a shorter Premier Support period and no Extended Support at all.

Release	Type	Premier Support	Notes
19c	Long Term Support	Through Dec 31, 2029 (Extended through Dec 31, 2032)	Previous LTS; still widely used on-premises and in the cloud
26ai	Long Term Support	Through Dec 31, 2031 (Extended TBD)	Current LTS
21c	Innovation	Through Jul 31, 2027	Not LTS; no Extended Support
23c/23ai	Replaced	—	Became 26ai via the October 2025 release update (23.26.0); no separate upgrade required

Both 19c and 26ai are genuinely current, supported LTS releases; 23c/23ai isn't a separate ongoing option, applying the October 2025 release update is what turns a 23ai database into 26ai. For monitoring purposes specifically, both 19c and 26ai targets are managed the same way, through the Oracle Management Agent and EM 24ai's Database plug-in; the agent architecture doesn't differ by which LTS release you're running.

The Management Agent is the bridge between a monitored target and the Oracle Management Service. Three distinct communication paths are involved:

1. Agent to OMS (push model). The agent initiates and maintains the connection itself over HTTPS, typically on port 4903, rather than OMS polling the agent. This is how collected metrics, alert notifications, target discovery data, and job results actually get uploaded.
2. OMS to agent (command execution). Reverse communication, established when needed, for deploying patches or plug-ins, executing jobs like RMAN backups or shell scripts, and kicking off diagnostic collections such as ADDM or AWR snapshots.
3. Target to agent (local monitoring). The agent runs on the same host as the target it's watching, database, ASM, listener, and communicates locally through OS-level commands, SQL*Net for database access, log scraping, and lsnrctl for the listener.

Agent and OMS authenticate each other with X.509 certificates, and all of this traffic is encrypted over HTTPS. Access control beyond that runs through target privileges and role-based access inside Enterprise Manager itself.

Walked through as an actual sequence, monitoring a database target looks like this: the agent wakes on its normal collection interval and gathers metrics, CPU, tablespace usage, whatever's configured; it packages that data and pushes it over HTTPS to the OMS upload port; the OMS stores it in the Management Repository, itself an Oracle database; and if a job is scheduled against that target, an RMAN backup, say, the OMS pushes the command down to the agent, which executes it locally and reports the result back. Nothing in that flow requires the OMS to reach out and poll the agent first, the agent-initiated push is what keeps the model working even through firewalls that only allow outbound connections from the monitored host.

Unrelated to the agent specifically, but worth having in view: Multi-Threaded Server was renamed Oracle Shared Server beginning with Oracle 9i. Additional background processes can still appear when certain features are in use, shared servers among them, along with job queues and replication.

Control Utility	Commands	Description	Last Supported Release
Oracle Names Control Utility	All commands	Oracle Names is no longer supported.	9.2
Listener Control Utility	DBSNMP_START DBSNMP_STATUS DBSNMP_STOP	These commands controlled the Oracle Intelligent Agent for use with Oracle Enterprise Manager. The Oracle Management Agent is controlled through Enterprise Manager itself today, not through listener control commands.	8.1
	SET CONNECT_TIMEOUT SHOW CONNECT_TIMEOUT	These commands specified how long the listener waited for a client's request after the transport connection was established.	8.1
	SET USE_PLUGANDPLAY SHOW USE_PLUGANDPLAY	These commands instructed the listener to register database information with an Oracle Names server.	8.1

The DBSNMP-prefixed commands are worth a second look, since the name still shows up today in a completely different context: DBSNMP is also the name of a default Oracle database account historically used by monitoring tools, including Enterprise Manager itself, to connect to a target database and collect metrics. The account's continued existence in current Oracle versions doesn't mean the old SNMP-based control commands listed above still function, they don't, they were retired at release 8.1 along with the Intelligent Agent architecture they controlled. Don't confuse a still-existing account name with control-utility commands that have been gone for decades.

After completing this module, you'll be able to:

1. Describe the Oracle Management Agent's role and how it communicates with Enterprise Manager
2. Deploy a Management Agent, including the Remote Agent option introduced in EM 24ai
3. Distinguish the Enterprise Manager Management Agent from the unrelated Select AI Agent and Private Agent Factory features in 26ai
4. Define alert thresholds and monitoring targets in Enterprise Manager Cloud Control

The next lesson provides an overview of the Enterprise Manager Cloud Control console itself.