MCP-GLPI: Integrate AI Assistants with GLPI via Model Context Protocol

EF-TECH is proud to announce mcp-glpi, an open-source server that implements the Model Context Protocol (MCP) to expose the entire GLPI ecosystem to AI assistants such as Claude, Cline, GitHub Copilot, and any other MCP-compatible client.

What is MCP-GLPI?

#

MCP (Model Context Protocol) is an open standard by Anthropic that allows AI assistants to connect to external systems securely and in a structured way. mcp-glpi is the bridge between this protocol and GLPI — the most widely used open-source IT service management (ITSM) platform.

With it, you can ask an AI assistant in natural language:

• “List open high-urgency tickets”
• “Create a ticket reporting that the database server is at 95% CPU usage”
• “Show me João’s computer data”
• “What is the total asset count in the inventory?”
• “Add a follow-up to ticket #42 stating the issue has been resolved”

All without leaving your favorite AI assistant.

Key Features

#

The server exposes 84 MCP tools and 9 resources organized into the following categories:

Ticket Management (Full ITIL)

#

Problems and Changes

#

• List, get, create, and update problems
• List, get, create, and update changes

Asset Management

#

• Computers: list, get (with software, network ports, connections, documents), create, update, and delete
• Software: list, get, and create
• Network Equipment: list, get, and create
• Printers: list, get, and create
• Monitors: list, get, and create
• Phones: list, get, and create

Knowledge Base, Contracts, Suppliers, and Locations

#

• List, get, and create knowledge base articles (with text search)
• List, get, and create contracts, suppliers, locations, and projects

Users, Groups, and Categories

#

• List users, search by login, create user
• List and create groups, add user to group
• List ticket categories
• List entities

Statistics and Introspection

#

• Ticket counts by status (with entity/period filter)
• Total assets by type
• Ticket distribution by status, category, technician, entity, or month
• Searchable field catalog
• Active session information

Resources

#

Beyond tools, the server exposes resources in the glpi:// format:

• glpi://tickets/open — open tickets
• glpi://tickets/recent — recent tickets
• glpi://problems/open — open problems
• glpi://changes/pending — pending changes
• glpi://computers — computers
• glpi://groups — groups
• glpi://categories — categories
• glpi://stats/tickets — ticket statistics
• glpi://stats/assets — asset statistics

Technology

#

mcp-glpi is built with TypeScript and runs on Bun, offering:

• OAuth2 authentication with support for password grant, client_credentials, and bearer token
• Unified HTTP layer with automatic reauthentication on 401, retry on 5xx/429, and configurable timeouts
• Multi-criteria search with RSQL filters (AND/OR/AND NOT/OR NOT), pagination, and fetch_all
• Dynamic field mapping — translations between field_id and names are cached with a 1-hour TTL
• Resolved foreign keys — details return both the ID and readable name without additional queries
• Runtime input validation with Zod — clear errors instead of GLPI failures
• Security annotations — readOnlyHint on query tools, destructiveHint on write/delete operations
• Docker image based on oven/bun:1-alpine, running as non-root user (bunuser:1001)

How to Use

#

With Claude Desktop / Claude Code

#

Add to your claude_desktop_config.json:

{
  "mcpServers": {
    "glpi": {
      "command": "bunx",
      "args": ["mcp-glpi"],
      "env": {
        "GLPI_URL": "https://glpi.your-company.com",
        "GLPI_AUTH_METHOD": "password",
        "GLPI_USERNAME": "your_username",
        "GLPI_PASSWORD": "your_password",
        "GLPI_CLIENT_ID": "your_client_id"
      }
    }
  }
}

With Docker

#

docker pull ghcr.io/eftechcombr/mcp-glpi:1.0

docker run -i --rm \
  -e GLPI_URL="https://glpi.your-company.com" \
  -e GLPI_AUTH_METHOD="password" \
  -e GLPI_USERNAME="your_username" \
  -e GLPI_PASSWORD="your_password" \
  -e GLPI_CLIENT_ID="your_client_id" \
  ghcr.io/eftechcombr/mcp-glpi:1.0

Prerequisites

#

1. A GLPI instance (version 10 or 11) with API access
2. An OAuth2 client configured in GLPI (Setup → General → OAuth2 Client)
3. An MCP-compatible AI assistant (Claude, Cline, Continue, etc.)

Why We Built This

#

At EF-TECH, we work with GLPI daily — whether in service desk operations for our clients or developing Docker images and Helm Charts for containerized deployment.

We noticed that AI assistants are becoming increasingly present in IT teams’ daily workflows, yet a direct connection between them and management tools was still missing. mcp-glpi fills exactly that gap: instead of manually copying and pasting information, the AI assistant queries and records data directly in GLPI.

Roadmap

#

The server is already functional and in use, but we have several plans for evolution:

• Support for change approval workflows
• Custom reporting tools
• Expanded asset coverage (network devices, batteries, etc.)
• Integration with more assistants and IDEs

Contribute

#

mcp-glpi is open-source under the MIT license and available on GitHub:

🔗 https://github.com/eftechcombr/mcp-glpi

Contributions are welcome! Issues, pull requests, and suggestions are open to the community.

Useful Links

#

• GitHub Repository
• NPM Package
• Model Context Protocol
• GLPI Project
• GLPI API Documentation

At EF-TECH, we specialize in GLPI, cloud computing, and IT infrastructure. We offer expert support for GLPI deployment, maintenance, and integration with modern tools. Contact us to learn how we can help your team.