GLPI Agent: 2026 install and configuration guide

Citation capsule

• "GLPI is deployed in 186+ countries by 11+ million users in 2026 (Source: GLPI Project, 2026)"
• "GLPI Agent 1.x is the direct successor to FusionInventory Agent 2.x — same codebase, drop-in replacement announced by Teclib (Source: GitHub glpi-agent, 2024)"
• "GLPI server stable releases 11.0.7 and 10.0.25 shipped in May 2026 and are compatible with agent 1.x (Source: GLPI Project, 2026)"

GLPI Agent is the tool most IT teams hit the moment they roll out GLPI (the open-source IT asset manager under GPL v3) past a handful of devices. The official docs are dense, the FusionInventory legacy confuses installers, and choices differ across Windows, macOS, and Linux.

This guide gives the verified commands, the supported-task table, and an honest diagnostic of the agent vs its alternatives — no marketing detour. To place GLPI in the broader ITAM landscape, see also our IT asset management software comparison, the IT asset management reference, the enterprise ITAM list, and the GLPI alternative SaaS 2026.

What GLPI Agent actually is

GLPI Agent is an inventory binary that collects hardware and installed software on a device, then pushes that data to a GLPI server over HTTPS. It runs as a scheduled task on Windows, a systemd service on Linux, and a launchd job on macOS. It replaces FusionInventory Agent 2.x — same Perl codebase, same XML formats, drop-in replacement announced in 2021.

Three things to keep in mind. First, the project is led by Teclib (a French publisher) and the GLPI community. Second, the code is GPL v2/v3 — free, modifiable, redistributable. Third, the 1.x branch is the only officially maintained line since 2023 — FusionInventory Agent is legacy.

Which version to install in 2026

The current stable is GLPI Agent 1.17 (May 2026). Nightly builds ship on nightly.glpi-project.org/glpi-agent/ and stable releases on GitHub. No 32-bit since 1.8 — x64 Windows only, arm64 + Intel macOS, x86_64 + aarch64 Linux.

Platform	Format	Official source
Windows	GLPI-Agent-1.17_x64.msi	github.com/glpi-project/glpi-agent/releases
macOS Intel	.pkg or .dmg	github.com/glpi-project/glpi-agent/releases
macOS Apple Silicon	.pkg or .dmg arm64	github.com/glpi-project/glpi-agent/releases
Debian/Ubuntu	.deb	github.com/glpi-project/glpi-agent/releases
RHEL/Fedora	.rpm	github.com/glpi-project/glpi-agent/releases
Snap	snap install glpi-agent	snapcraft.io

The target GLPI server must run 10.0.x or 11.0.x. Latest stable releases are 10.0.25 and 11.0.7 (May 2026, official note).

Install GLPI Agent on Windows

Windows installs use the x64 MSI. Two modes — interactive (double-click) or silent (GPO/SCCM/Intune deploy).

Silent install with server URL (recommended for fleets):

msiexec /quiet /i GLPI-Agent-1.17_x64.msi ^
  SERVER="https://glpi.example.com/marketplace/glpiinventory/" ^
  RUNNOW=1

Key MSI variables:

• SERVER=<url> — inventory push URL (server-side glpiinventory plugin)
• RUNNOW=1 — triggers an inventory immediately after install
• TAG=<label> — tag to group devices
• HTTPD_TRUST=127.0.0.1 — IPs allowed to query the local API

The agent registers as a scheduled task under Microsoft\GLPI-Agent running as SYSTEM. Config file: %ProgramData%\GLPI-Agent\agent.cfg.

Install GLPI Agent on macOS

On Apple Silicon (M1/M2/M3/M4), use the arm64 .pkg. The agent installs under /Applications/GLPI-Agent.app and runs via launchd. The binary is Developer ID signed but not always notarized — Gatekeeper may block; right-click > Open to bypass.

sudo installer -pkg GLPI-Agent-1.17_arm64.pkg -target /
sudo /Applications/GLPI-Agent.app/Contents/MacOS/glpi-agent \
  --server=https://glpi.example.com/marketplace/glpiinventory/ \
  --tag=production

Config file lives at /Applications/GLPI-Agent.app/Contents/Resources/etc/agent.cfg.

Install GLPI Agent on Linux

Debian/Ubuntu:

sudo dpkg -i glpi-agent_1.17_all.deb
sudo apt-get install -f
sudo systemctl enable --now glpi-agent

RHEL/Fedora:

sudo dnf install glpi-agent-1.17.noarch.rpm
sudo systemctl enable --now glpi-agent

Default config is /etc/glpi-agent/agent.cfg or /etc/glpi-agent/conf.d/local.cfg. Edit server = https://glpi.example.com/marketplace/glpiinventory/, then restart the service.

The 8 tasks GLPI Agent can run

The agent exposes 8 official tasks, each enabled individually via config.

Task	Role
inventory	Hardware + software inventory of the local device
netdiscovery	SNMP discovery of IP ranges — detects switches, printers, routers
netinventory	Detailed SNMP inventory of discovered equipment
esx	Inventory of VMware ESXi hosts via vSphere API
deploy	Package deployment (MSI, EXE, scripts) — orchestrated by the server
wakeonlan	Send Wake-on-LAN magic packets
collect	Custom data collection (files, registry keys, command output)
remoteinventory	Remote inventory over SSH/WinRM without an installed agent

Default schedule: PROLOG_FREQ = 3600 seconds (1 hour) for the prolog; full inventory is typically daily. All parameters live in agent.cfg.

GLPI Agent vs FusionInventory Agent — which to pick

Pick GLPI Agent 1.x. Always. FusionInventory Agent 2.x has been legacy since 2021 — no new features, no official support, no signed packages. Migration is drop-in: uninstall FusionInventory, install GLPI Agent, keep the same server. Previous inventories stay intact on the GLPI side because the XML formats are compatible.

The only edge case where FusionInventory 2.x still matters: a specific plugin (rare) that hasn't been ported. Check the GLPI forum before migrating.

GLPI Agent vs other inventory tools

Tool	Model	Cost	Strength	Limit
GLPI Agent	Perl agent per device + self-hosted GLPI server	Free (GPL v3)	Open source, mature ecosystem, cross-OS	Bundled Perl runtime, CPU peaks during scans
OCS Inventory NG Agent	Per-device agent + OCS server	Free (GPL v2)	Lightweight network footprint, pure inventory	Aging UI, smaller community
Lansweeper	Agentless scan (WMI/SSH/SNMP) + optional agent	€199/mo (Starter 2000 assets)	No agent on devices by default	Inventory less rich without agent, cost scales with asset count
Snipe-IT	Manual entry or integrations	$39.99/mo cloud / free self-hosted	Excellent for physical traceability	No native automated inventory
Microsoft Intune	MDM agent bundled with Microsoft 365	Included in some M365 licenses	Native Entra ID integration, compliance	Locks you into the Microsoft cloud
Action1	Cloud SaaS agent	Free ≤200 endpoints, then $4/endpoint/month (Growth)	Native patching	No self-hosting, narrow ITAM scope

One Rust agent vs the GLPI Perl stack

One Rust agent, < 1% CPU. Where GLPI Agent (Perl) has to bundle its runtime — ~25 MB on Windows disk — and runs in peaks during scans, sobrii ships a single signed Rust binary covering inventory, per-app energy measurement, zombie-app detection, and WebRTC remote control. Measured footprint: < 1% CPU continuously on Windows and macOS.

For an IT team stacking GLPI Agent + an MDM (Intune) + an EDR (CrowdStrike) + an RMM (Atera) + a DEX tool (ControlUp), the stack hits 4 to 5 agents per device. Fewer agents = smaller attack surface, less battery drain, less support coordination.

Deploying GLPI Agent at scale (500+ endpoints)

Scaling GLPI Agent across 500+ Windows devices requires orchestration via Group Policy, Intune, or SCCM. The typical deployment timeline is 3–4 weeks from planning to 95% coverage. Create an OU for GLPI Agent targets, use the MSI with /quiet /server=<url> /tag=<label>, and validate inventory push via the GLPI server UI's dashboard. Monitor initial CPU load on week one — peaks of 20–30% during the first scan are normal as WMI enumerates hardware. By week two, scans stabilize at 5–10% during the hourly prolog and daily full inventory. Uninstall is clean via Control Panel or msiexec /quiet /x {GUID} — no registry cruft.

GLPI Agent maintenance and troubleshooting

After deployment, GLPI Agent requires minimal hand-holding. Most issues stem from three causes: (1) firewall blocking HTTPS egress to the GLPI server URL, (2) stale agent.cfg on a subset of devices pointing to an old server, and (3) certificate validation failures when using self-signed server HTTPS. For each, the fix is quick: verify the firewall rule, push a corrected MSI via Intune, or disable certificate pinning in agent.cfg (verify-certs = false — use only for testing). Check agent logs at %ProgramData%\GLPI-Agent\glpi-agent.log on Windows and /var/log/glpi-agent/ on Linux. Errors typically appear within the first 100 lines.

See how the single sobrii agent covers inventory and remote control in one binary.

The 4th lifecycle decision no inventory agent computes

sobrii adds a 4th lifecycle decision: reallocate. GLPI, Lansweeper, and Snipe-IT offer 3 paths on an aging device — keep, repair, replace. sobrii computes 4 options per device — upgrade, repair, reallocate to the next employee, replace — with cost and CO₂ for each. The reallocate branch extends average service life by 12-18 months and halves per-device embodied carbon.

GLPI Agent gathers the raw data (CPU, RAM, age, disk SMART state). The GLPI user decides afterward — with no numbered recommendation. sobrii surfaces the recommended decision with TCO and CO₂ math per scenario.

kWh-per-employee measurement — the data the Perl agent doesn't expose

sobrii measures kWh per employee, not per site. The Rust agent captures real consumption (CPU/GPU/screen/battery) second-by-second, then applies the regional grid emission factor (Ember/EPA eGRID) to produce kg-CO₂/employee/month — exportable directly to CSRD ESRS E1.

GLPI Agent doesn't measure power consumption per device — it lists hardware. To produce a CSRD report, you have to cross GLPI with a third-party measurement tool (or estimate by category). sobrii rolls the measure → emission factor → ESRS export chain into one pipeline.

WebRTC remote desktop bundled — no TeamViewer add-on

Remote control is in the plan, not an add-on. No need to buy TeamViewer or AnyDesk on top: sobrii ships a built-in WebRTC remote-desktop module (peer-to-peer, no external relay, multi-screen, auto-reconnect). 200-device benchmark: TeamViewer Business fully-loaded median spend ≈ $1,020/month → sobrii: $0/month (bundled).

GLPI Agent can trigger wake-on-LAN, but doesn't offer interactive control. Most IT teams running GLPI bolt a TeamViewer/AnyDesk license alongside — a recurring overspend you can avoid.

100% bilingual FR/EN — native, not 70%-translated

sobrii is 100% bilingual FR/EN at the product core. Every label, every CSRD/ISO 27001 report, every export is rendered in the user's language — not a 70%-translated glossary. Reference customer: Montpellier Metropolitan Area (3M residents, multi-site fleet). sobrii is one of the rare ITAM SaaS designed in France with FR/EN parity from v1.

GLPI itself is bilingual (it's a French project), but parts of the extended documentation, third-party plugins, and integrations remain EN-only.

FAQ

What is the latest stable version of GLPI Agent

The current stable is GLPI Agent 1.17 (May 2026), available from the GitHub releases page. Nightly builds publish on nightly.glpi-project.org/glpi-agent/ for early testing. No 32-bit version since 1.8.

Does GLPI Agent run on Apple Silicon

Yes — native arm64 .pkg and .dmg packages ship for M1/M2/M3/M4 on the recent 1.x branch. No Rosetta emulation required. Intel builds remain maintained in parallel.

Can I migrate from FusionInventory Agent to GLPI Agent without losing history

Yes. GLPI Agent 1.x is the direct successor to FusionInventory Agent 2.x — same codebase, same XML format. Uninstall FusionInventory, install GLPI Agent, point at the same server. Past inventories stay intact. FusionInventory 2.x has been legacy since 2021 — don't delay migration.

Does GLPI Agent collect installed software or only hardware

Both. The inventory task gathers hardware (CPU, RAM, disks, BIOS, screens, battery) and software (installed apps via WMI on Windows, dpkg/rpm on Linux, /Applications on macOS). Windows patches, user accounts, and connected printers are also reported.

How heavy is GLPI Agent on a device

The agent idles between scheduled scans (hourly prolog by default, daily full inventory typical). During a scan, CPU peaks 15-30% for a few seconds — that's the Perl + WMI cost. Idle: < 50 MB RAM, 0% CPU. The peaks hurt lightly-spec'd laptop fleets.

Does GLPI Agent work without a GLPI server

Partially. The inventory task can produce a local XML (option --local=<path>) without pushing to a server — useful for one-off audits. But the full model (deployment, scheduling, history) assumes an active GLPI 10.x/11.x server. For SaaS use without self-hosting, look at cloud alternatives.

What's the difference between netdiscovery and netinventory

netdiscovery sweeps IP ranges via SNMP/ICMP to detect which devices respond (switches, printers, routers, servers). netinventory then queries each detected device for detail (serial number, BIOS, ports, toner level). Both are agentless from the target's perspective — only the agent host runs Perl.

How do I validate that GLPI Agent is reporting inventory successfully?

Check the GLPI server dashboard — navigate to Assets > Devices and verify that your agents appear and inventory dates are recent (within the last 24 hours). On the agent, check the log file for successful transmission messages like "Successfully sent prolog." Cross-check with the GLPI server's Request->Requests view to see unprocessed queues — if messages are backing up, it usually signals a server-side plugin issue or API authentication failure.

Can GLPI Agent coexist with Lansweeper or other inventory agents?

Yes, GLPI Agent runs alongside other agents. Conflicts are rare because agents use separate scheduled-task names and config paths. However, running multiple agents simultaneously increases CPU load, especially during scans. On a fleet where both GLPI and Lansweeper are deployed, stagger scan times via the PROLOG_FREQ setting to avoid sync peaks — set GLPI to hourly and Lansweeper to 90 minutes, for example.

What happens if GLPI Agent loses network connectivity?

The agent gracefully handles disconnects. Inventory data is queued locally at %ProgramData%\GLPI-Agent\var\ on Windows. Once the agent regains connectivity, queued inventories upload automatically on the next scheduled prolog. No data loss occurs unless the agent is uninstalled before reconnection. For fleets in remote offices with intermittent WAN connectivity, increase PROLOG_FREQ to 2-4 hours to reduce retry storms.