The GLPI agent is the data pipeline that feeds your CMDB. Without it, GLPI is just an empty database with a ticket form. With it, you get automatic hardware and software inventory for every machine that has the agent installed. The practical challenge is not whether to deploy it — it is how to roll it out to hundreds or thousands of endpoints without losing a weekend.
Deployment methods
Group Policy (Windows environments) — Package the GLPI agent MSI with your desired configuration parameters (server URL, scan interval, enabled tasks) and deploy it through a GPO linked to the appropriate OUs. The agent installs silently and runs as a Windows service. Machines pick up the policy on their next group policy refresh — typically within 90 minutes, or immediately if you force a gpupdate.
SCCM / MECM — If you already use Microsoft Endpoint Configuration Manager, deploy the agent as a standard application package. This gives you deployment monitoring, success/failure reporting, and the ability to target specific device collections. Useful when you want to stage the rollout — pilot group first, then department by department.
Ansible or scripts (Linux and mixed environments) — For Linux servers and mixed fleets, write an Ansible playbook or a shell script that installs the agent package, drops the configuration file, and starts the service. The GLPI agent ships as .deb, .rpm, and .pkg packages, so your package manager handles dependencies.
Manual install — Fine for a handful of machines. Not viable at scale. If you are still installing agents by walking to desks with a USB stick, any of the methods above will save you days per year.
Configuration essentials
The agent configuration file (agent.cfg on Linux, registry keys or agent.cfg on Windows) needs a few things set correctly:
- Server URL — the address of your GLPI instance, including the inventory endpoint path (
/front/inventory.phpor the newer/Inventoryendpoint in GLPI 10.0.x and later) - Scan interval — how often the agent runs a full inventory. Daily (86400 seconds) is the standard choice. More frequent scans add server load without much benefit unless you have high device churn.
- Enabled tasks —
inventoryis the baseline. Addnetdiscoveryandnetinventoryonly on designated scanner agents, not on every endpoint. - Tag — optional but useful. Tags let you auto-assign incoming assets to specific GLPI entities based on the tag value. Set this to match your entity structure if you run a multi-entity GLPI.
What happens after rollout
Within 24-48 hours of deployment, assets start appearing in GLPI under Administration > Inventory. Each agent submission creates or updates a computer record with hardware specs (CPU, RAM, disks, monitors, network interfaces), a full software list with version numbers, operating system details, last boot time, and network configuration (IPs, MACs, gateways).
The first run is the noisiest — expect a spike in new asset records. After that, the agent only reports changes, so subsequent submissions generate less data and less server load.
Keeping agents healthy
Version drift — When you upgrade GLPI, check whether the agent version needs to match. Major GLPI upgrades (for example 10.0.x to 11.x) often require a corresponding agent update. Running a mismatched version can silently drop inventory fields.
Offline agents — Monitor the "last inventory" date in GLPI. An agent that has not reported in 30+ days usually means the machine is powered off, reimaged, or the agent service crashed. The built-in agent management view under Administration > GLPI Agent shows last contact times at a glance.
Duplicate assets — This happens when machines get reimaged without clearing the agent device ID. Configure GLPI's asset reconciliation rules to match on serial number or MAC address rather than the agent-generated UUID alone. That way a reimaged machine updates its existing record instead of creating a ghost entry.
A well-deployed agent fleet turns GLPI from a manual data-entry tool into a live infrastructure map. The initial deployment takes a few hours of preparation; the payoff is years of accurate, automatic inventory.