mcp-glpi

mcp-glpi

#

mcp-glpi v3.4 is a Model Context Protocol (MCP) server that connects MCP-compatible AI assistants to GLPI. It is based on the upstream GMS64260/mcp-glpi project and exposes standardized tools for querying and operating tickets, assets, problems, changes, and other GLPI resources.

The project uses the stable High-Level API v2.3 path, compatible with GLPI 11 and GLPI 12. EF-TECH can support use-case evaluation, deployment, and operation of the integration according to the organization’s environment and requirements.

Coverage for service desk and operations

#

mcp-glpi provides 88 MCP tools across the main operational areas of GLPI:

• Tickets, followups, tasks, solutions, validations, and SLA tracking
• Problems and changes
• Computers, software, network equipment, printers, monitors, and phones
• Knowledge base, contracts, suppliers, locations, and projects
• Users, groups, categories, entities, and documents
• Multi-criteria search, counts, and statistics

Ticket tools include chronological timelines, user or group assignment, ticket linking, validation requests, and task lifecycle management with time tracking. This lets AI assistants query operational context and perform GLPI actions while respecting the permissions and configuration of the GLPI instance.

Search and operational visibility

#

Multi-criteria search accepts RSQL filters with AND, OR, AND NOT, and OR NOT, together with field selection, pagination, full-result retrieval, and a dedicated count probe. Dynamic searchable-field mapping helps keep the integration valid across GLPI versions. Ticket search provides direct parameters for status, assignee, requester, and date ranges.

The server can report ticket statistics by status, category, technician, entity, or month, asset totals by type, and tickets past their SLA deadline. Detail responses can resolve foreign keys and return both the related ID and readable name.

OAuth2 authentication and safety signals

#

Tools include MCP safety annotations to distinguish read and write operations. Listing, get, search, count, and statistics tools receive readOnlyHint; update, delete, set, and assignment operations receive destructiveHint. Ticket tools also validate runtime inputs with Zod and return structured InvalidParams errors when parameters are invalid.

mcp-glpi v3 uses OAuth2 with password, client_credentials, or a bearer token. The v2 API app_token and user_token authentication is not supported in v3. An OAuth2 client with the appropriate grant must be configured in GLPI; credentials are supplied through environment variables or the HTTP session configuration. See the project README for the complete variable table and requirements for each method.

Transport: stdio or Streamable HTTP

#

With stdio transport, the MCP client starts the server as a subprocess and exchanges messages through stdin/stdout. This is suitable for local integrations or a container started by the client. With Docker, keep -i to preserve stdin and do not use -t, because a TTY can corrupt the protocol stream:

docker pull ghcr.io/eftechcombr/mcp-glpi:latest
docker run -i --rm --env-file /absolute/path/.env \
  ghcr.io/eftechcombr/mcp-glpi:latest

With Streamable HTTP transport, the server accepts network connections. Set MCP_TRANSPORT=http and publish the service port:

docker run -d --name mcp-glpi-http -p 3000:3000 \
  -e MCP_TRANSPORT=http -e PORT=3000 \
  ghcr.io/eftechcombr/mcp-glpi:latest

The default MCP endpoint is /mcp; /health and / provide health endpoints. When using a reverse proxy, the /mcp path must forward streaming without SSE buffering.

Streamable HTTP supports two operating models:

• Single-tenant: GLPI_URL and the authentication variables define one instance for sessions.
• Multi-tenant: leave GLPI_URL unset and let each session configure its own connection with glpi_configure_connection. Credentials remain in memory and are removed when the session ends or expires.

In multi-tenant mode, the client checks glpi_connection_status, calls glpi_configure_connection with the URL, OAuth2 method, and matching credentials, and then uses the tools normally.

How EF-TECH can help

#

EF-TECH can support AI use-case assessment for service desk, GLPI and MCP client configuration, containerized execution, and service operations. The scope is defined according to the environment’s architecture, permissions, and security requirements.

See the mcp-glpi repository on GitHub and the upstream project for complete configuration, the tool catalogue, and development details.

Want to connect AI to your company’s GLPI? Contact us to speak with our team.