Most IT teams start their CMDB the same way: someone opens a spreadsheet and starts typing serial numbers. It works for 50 machines. It falls apart at 500. The GLPI agent (the project formerly known as FusionInventory) solves this by scanning your network and populating asset records automatically — no clipboard walks required.
How the GLPI agent discovers devices
The GLPI agent runs in two modes, and you will likely use both.
Agent-based inventory means installing the agent directly on endpoints — Windows, Linux, or macOS. The agent collects hardware specs, installed software, connected peripherals, and network configuration, then sends a structured XML/JSON payload to the GLPI server on a schedule you define. This is the most detailed collection method because the agent has local access to everything the OS knows about itself.
Agentless (SNMP) network discovery works differently. You configure one or more agents to act as network scanners. They sweep IP ranges using SNMP queries and collect data from any device that responds — switches, routers, printers, access points, IP phones, UPS units. These devices never get an agent installed on them; instead, the scanner reads their SNMP MIBs remotely.
What SNMP discovery actually collects
The data you get from an SNMP-discovered device depends on what the device exposes through its MIBs, but typically includes:
- IP address and MAC address for each interface
- Device hostname and system description (sysDescr)
- Serial number and model (when the vendor populates these OIDs)
- Firmware version
- Port status and connected device mapping (via LLDP/CDP)
- Uptime and interface traffic counters
For managed switches, you also get port-to-MAC mappings, which lets GLPI build a network topology — showing which endpoint is plugged into which switch port. That alone is worth the setup effort when you are troubleshooting connectivity issues.
Setting up network scans
Configuration happens in two places: the agent configuration file and the GLPI web interface.
On the agent side, enable the netdiscovery and netinventory tasks. Point the agent at your GLPI server URL and set the scan interval. Hourly is common for discovery; daily works fine for full inventory pulls.
In GLPI, go to Administration > GLPI Agent and define your IP ranges. For each range, assign SNMP credentials (community strings for v1/v2c, or user/auth/priv for SNMPv3). A practical tip: create separate credential sets for different device families. Your Cisco switches probably use a different community string than your HP printers.
Getting SNMPv3 right
If your security policy requires SNMPv3 — and it should for anything crossing untrusted network segments — you need to configure the authentication protocol (MD5 or SHA), privacy protocol (DES or AES), and the corresponding passwords in GLPI’s credential store. The agent will use these when querying each device. SNMPv3 takes more setup, but it encrypts the traffic and prevents community string sniffing.
Agent-based vs. agentless: when to use which
Use agent-based collection for anything you manage directly — workstations, servers, laptops. The data is richer (full software inventory, local user accounts, disk health), and you get it regardless of network topology.
Use SNMP discovery for infrastructure devices that cannot run an agent — switches, firewalls, printers, IP cameras. Also use it as a verification layer: if a device shows up in SNMP discovery but has no agent record, something is unmanaged that should not be.
Running both methods on the same network gives you the most complete picture. The agent handles depth; SNMP handles breadth. Together, they turn GLPI from a manual database into a live inventory that reflects what is actually connected to your network right now.