GLPI MCP Server

Connect AI assistants (Claude, ChatGPT, Copilot) to your GLPI IT Service Management instance via the Model Context Protocol (MCP).

First MCP Server for GLPI — Manage tickets, search assets, and automate IT operations through natural language.

Features

List Tickets — View open, assigned, waiting, or all tickets
Get Ticket Details — Full ticket info with description and followup timeline
Create Tickets — Open incidents or service requests via AI
Update Tickets — 🆕 Change status, priority, or attribution
Add Followups & Solutions — 🆕 Reply to or solve tickets directly
Manage Tasks — 🆕 Create and view technical tasks for tickets
ITIL Processes — 🆕 Create Changes and Problems
Search Everything — Find computers, users, software, printers, and more

Supported Versions

| GLPI Version | API | Status | |-------------|-----|--------| | 10.0.x | Legacy REST API (/apirest.php) | ✅ Supported | | 11.0.x | HLAPI (/api.php) | ✅ Supported |

Quick Start

1. Prerequisites

Node.js 18+
A GLPI instance with the REST API enabled
An API User Token or username/password

2. Setup GLPI API Access

In your GLPI instance:

Go to Setup > General > API
Enable REST API (set to YES)
Enable authentication method (Credentials and/or External Token)
Note the API URL shown at the top
Get your User Token from your user profile > Remote Access Keys > Regenerate

3. Install & Build

git clone https://github.com/your-user/glpi-mcp-server.git
cd glpi-mcp-server
npm install
npm run build

4. Configure Claude Desktop

Add to your claude_desktop_config.json:

macOS: ~/Library/Application Support/Claude/claude_desktop_config.json Windows: %APPDATA%\Claude\claude_desktop_config.json Linux: ~/.config/Claude/claude_desktop_config.json

{
  "mcpServers": {
    "glpi": {
      "command": "node",
      "args": ["/absolute/path/to/glpi-mcp-server/dist/index.js"],
      "env": {
        "GLPI_URL": "https://your-glpi-instance.com",
        "GLPI_USER_TOKEN": "your_user_token_here",
        "GLPI_APP_TOKEN": "your_app_token_here"
      }
    }
  }
}

5. Restart Claude Desktop

After saving the config, restart Claude Desktop. You should see the GLPI tools available in the tools menu.

Environment Variables

| Variable | Required | Description | |----------|----------|-------------| | GLPI_URL | Yes | Base URL of your GLPI instance. Do not include /api.php or /apirest.php unless you intentionally want to override auto-detection. | | GLPI_API_VERSION | No | API version. Defaults to 10; set 11 to enable HLAPI routing and OAuth2 authentication. | | GLPI_USER_TOKEN | v10 only* | User API token from the user profile. Preferred for GLPI 10 legacy REST auth. | | GLPI_USERNAME | v11 required | GLPI username. Also valid for GLPI 10 when used with GLPI_PASSWORD. | | GLPI_PASSWORD | v11 required | GLPI password. Also valid for GLPI 10 when used with GLPI_USERNAME. | | GLPI_APP_TOKEN | No | App token from API client config. | | GLPI_OAUTH_CLIENT_ID | v11 required | OAuth2 Client ID for GLPI 11 HLAPI. | | GLPI_OAUTH_CLIENT_SECRET | v11 required | OAuth2 Client Secret for GLPI 11 HLAPI. |

\ For GLPI 10, either GLPI_USER_TOKEN or both GLPI_USERNAME + GLPI_PASSWORD are required. \* For GLPI 11, GLPI_USERNAME, GLPI_PASSWORD, GLPI_OAUTH_CLIENT_ID, and GLPI_OAUTH_CLIENT_SECRET are required.

Tools Reference

`glpi_list_tickets`

List tickets with optional status filter.

"Show me all open tickets"
"List waiting tickets, max 10"
"What tickets are assigned right now?"

`glpi_get_ticket`

Get complete details of a specific ticket.

"Show me ticket #42"
"What's the status of ticket 150?"
"Get the timeline for ticket 99"

`glpi_create_ticket`

Create a new incident or service request.

"Open a ticket: WiFi not working in meeting room 3"
"Create a request for a new monitor for João Silva"

`glpi_add_followup`

Add a followup message to an existing ticket.

"Reply to ticket 42: the router has been restarted"
"Add a private note to ticket 100 saying we're waiting for parts"

`glpi_search`

Search across any GLPI item type. Now supports advanced multi-criteria searches with AND/OR logic!

"Find all computers where location is Finance and OS is Windows"
"Search for user named Maria"
"List HP printers"
"Find software named Chrome"

Supported item types: Ticket, Computer, User, Software, NetworkEquipment, Monitor, Printer

`glpi_list_search_options`

Discover all possible search fields and their numeric IDs for any GLPI item type. Essential for advanced searches where fields aren't mapped by default (like Antivirus, installed software, etc.).

`glpi_update_ticket` 🆕

Update an existing ticket's name, content, status, priority, etc.

`glpi_add_solution` 🆕

Add a solution to a ticket and mark it as solved.

`glpi_add_task` 🆕

Add a technical task to a ticket with state, planning, and duration.

`glpi_get_ticket_tasks` 🆕

List all tasks associated with a ticket.

`glpi_create_change` 🆕

Create an ITIL Change Request.

`glpi_create_problem` 🆕

Create an ITIL Problem report.

`glpi_request_validation` 🆕

Request validation for a ticket from a supervisor or manager.

`glpi_get_ticket_validations` 🆕

List all validation requests associated with a ticket.

`glpi_answer_validation` 🆕

Approve or Refuse an existing ticket validation request.

`glpi_list_forms` 🆕

Discover available native GLPI 11 forms in the service catalog.

`glpi_get_form_details` 🆕

Retrieve a native form's questions and sections.

`glpi_submit_form` 🆕

Submit structured answers to a native GLPI 11 form.

Claude Code Configuration

For use with Claude Code, create a .mcp.json in your project root.

Example for GLPI 11 (OAuth2): ``json { "mcpServers": { "glpi": { "command": "node", "args": ["/absolute/path/to/glpi-mcp-server/dist/index.js"], "env": { "GLPI_URL": "https://your-glpi-instance.com", "GLPI_API_VERSION": "11", "GLPI_USERNAME": "your_username", "GLPI_PASSWORD": "your_password", "GLPI_OAUTH_CLIENT_ID": "your_client_id", "GLPI_OAUTH_CLIENT_SECRET": "your_client_secret" } } } } ``

Example for GLPI 10 (Legacy REST): ``json { "mcpServers": { "glpi": { "command": "node", "args": ["/absolute/path/to/glpi-mcp-server/dist/index.js"], "env": { "GLPI_URL": "https://your-glpi-instance.com", "GLPI_USER_TOKEN": "your_user_token_here" } } } } ``

Development

# Watch mode (recompile on changes)
npm run dev

# Test with MCP Inspector
npx @modelcontextprotocol/inspector node dist/index.js

Project Structure

glpi-mcp-server/
├── src/
│   ├── index.ts                  # Entry point, server setup
│   ├── types.ts                  # TypeScript types and constants
│   ├── services/
│   │   ├── glpi-client.ts        # GLPI REST API client
│   │   └── formatting.ts         # Response formatting utilities
│   └── tools/
│       ├── add-followup.ts       # glpi_add_followup
│       ├── add-solution.ts       # glpi_add_solution
│       ├── add-task.ts           # glpi_add_task
│       ├── answer-validation.ts  # glpi_answer_validation
│       ├── create-change.ts      # glpi_create_change
│       ├── create-problem.ts     # glpi_create_problem
│       ├── create-ticket.ts      # glpi_create_ticket
│       ├── get-form-details.ts   # glpi_get_form_details
│       ├── get-ticket.ts         # glpi_get_ticket
│       ├── get-ticket-tasks.ts   # glpi_get_ticket_tasks
│       ├── get-ticket-validations.ts # glpi_get_ticket_validations
│       ├── list-forms.ts         # glpi_list_forms
│       ├── list-search-options.ts # glpi_list_search_options
│       ├── list-tickets.ts       # glpi_list_tickets
│       ├── request-validation.ts # glpi_request_validation
│       ├── search.ts             # glpi_search (v2 Multi-Criteria)
│       ├── submit-form.ts        # glpi_submit_form
│       └── update-ticket.ts      # glpi_update_ticket
├── jest.config.js
├── .env.example
├── package.json
├── tsconfig.json
└── README.md

Knowledge Base & Guidelines

This section serves as the continuous knowledge base for project progress, maintainability, and testing.

Personas

IT Administrator / Analyst (End-User): Uses AI assistants (like Claude) connected via this MCP server to manage tickets, search for assets, and automate daily IT operations without opening the GLPI interface.
MCP Server Developer / Contributor: TypeScript developer responsible for maintaining the GLPI integration, adding new endpoints, ensuring API backward/forward compatibility (e.g., GLPI 10 REST to GLPI 11 HLAPI), and optimizing performance.
AI Agent (System Persona): The LLM client connecting via MCP that acts as the intelligent middleman. It needs structured, clear, and comprehensive descriptions of tools and schemas to effectively parse user intent into GLPI actions.

Rules

Code & Architecture: Maintain the clear separation of concerns. Keep src/services/ for API communication and data formatting, and src/tools/ for the MCP tool definitions. All new code must be written in TypeScript.
API Interactions: Always respect GLPI API limits and handle authentication gracefully. Validate GLPI_USER_TOKEN and application tokens securely.
Type Safety: Strictly use TypeScript interfaces and Zod schemas (from @modelcontextprotocol/sdk) to validate all inputs from the AI model before passing them to the GLPI client.
Error Handling: Format errors in a way that the AI Agent can read and relay back to the user clearly (e.g., "Missing permissions" instead of raw 403 stack traces).
Stability: Ensure the server is stateless and can quickly recover or handle multiple concurrent agent requests.
Documentation Sync: Every sensitive or architectural code change, new entity mapping, or workflow alteration MUST be immediately reflected inside this README.md file to keep it as the reliable single source of truth for the project.
Roadmap Validation: The AI Assistant MUST proactively check the Roadmap section at the start of every session and update it whenever a feature moves from "In Progress" to "Completed". All new features discussed must be added to the roadmap before implementation begins.
Automated Testing: New or changed backend components, formatting utilities, and tool abstractions must be covered by automated tests (Jest). See TESTS.md for full testing rules, routines, and workflows.

Workflows

1. Developing a New Tool

Define the Schema: Identify the GLPI endpoints needed. Create Zod schemas for the inputs.
Implement the Service: Add the corresponding API call method to src/services/glpi-client.ts.
Create the Tool Module: Create a new file in src/tools/ defining the tool's name, description, and execution logic.
Register: Import and register the new tool in src/index.ts.
Compile: Run npm run build to update the dist/ directory.

2. Local Testing Workflow

Watch Mode: Run npm run dev to automatically recompile TypeScript on save.
MCP Inspector: Use the inspector to test the server in isolation (without Claude Desktop):

   npx @modelcontextprotocol/inspector node dist/index.js

Configuration: Supply required environment variables within the inspector's UI to run tests against a staging or test GLPI environment. For GLPI 10, use GLPI_URL plus GLPI_USER_TOKEN or GLPI_USERNAME + GLPI_PASSWORD. For GLPI 11, use GLPI_URL, GLPI_API_VERSION=11, GLPI_USERNAME, GLPI_PASSWORD, GLPI_OAUTH_CLIENT_ID, and GLPI_OAUTH_CLIENT_SECRET.
Validation: Validate successful creation, retrieval, edge cases (e.g., invalid ticket IDs), and proper error returns.

Production Notes & Tips

WAF / Firewall Bypass: Strict corporate firewalls often strip custom HTTP headers like App-Token or Session-Token, causing 400 ERROR_WRONG_APP_TOKEN_PARAMETER errors. To ensure maximum compatibility, this MCP server automatically injects tokens into the URL query string (?app_token=...) instead of HTTP headers for GLPI 10 legacy authentication.
Group Search: GLPI uses hierarchical group names (e.g., TECH_DEPT > INFRA > NETWORKS). When searching tickets by group, always use the contains search type with the last part of the name (the "leaf") for best results.
Assignment Required: In some glpi11 environments, you must assign a technician to a ticket before the API allows registering a solution. If you get a 400 error when solving, check if the ticket has an assigned technician.
Escalade Plugin: If your instance uses the Escalade plugin, you can search for escalated tickets using field ID 1881 ("Grupo afetado pela escalada").
API Versions: GLPI_API_VERSION="11" enables HLAPI routing through /api.php with path patterns like /Assistance/Ticket. Without this variable, the server defaults to GLPI 10 legacy REST routing through /apirest.php.

3. Automated Testing Routine (Jest)

For all instructions related to writing tests, running tests, and test execution rules, please refer directly to the TESTS.md document.

Skills & Competencies Required

TypeScript / Node.js: Advanced proficiency for robust server-side development.
GLPI Architecture: Deep understanding of GLPI's Itemtypes (Ticket, Computer, User, Software), ITIL concepts, and legacy REST API payload structures.
Model Context Protocol (MCP): Knowledge of building and debugging MCP servers using the official @modelcontextprotocol/sdk.
System Integration: Ability to design safe, predictable integrations between unpredictable LLMs and strict database-backed systems.

Roadmap

Core v0.2.0 (Completed ✅)

[x] Cross-Version Support: Dynamic routing for glpi10 and glpi11.
[x] Full ITSM Cycle: Ticket Solution, Tasks, Changes, and Problems.
[x] Advanced Search: Multi-criteria AND/OR search with field ID mapping.
[x] Escalade Plugin Support: Field 1881 mapping for glpi11.

Enterprise ITSM (v0.3.0) ✅

[x] Approval Workflow (Validations):
glpi_request_validation: Send approval requests to Managers.
glpi_answer_validation: Approve or Reject via Chat.
glpi_get_ticket_validations: Monitor status of approvals.
[x] Native Forms (GLPI 11):
glpi_list_forms: Discovery of available service catalogs.
glpi_get_form_details: Retrieve questions and form configuration.
glpi_submit_form: Submit structured requests (Tickets/Changes).

Enterprise ITSM (Development 🏗️)

[ ] Phase 3: SLA/OLA Management:
glpi_check_prazos: Real-time monitoring of TTO and TTR countdowns.

🚀 Enterprise Blueprint: Approval Workflow

To achieve the Full ITSM Cycle, we will focus first on the Validation Engine. This allows the AI to act as a governance layer between the user and the technical execution.

Implementation Strategy:

Tool Logic: The AI must check for existing validations before requesting a new one.
Status Sync: If a validation is refused, the AI should optionally move the Ticket to a Waiting or Closed status automatically.
Escalation: Integration with the Escalade plugin to ensure validations follow the correct hierarchical chain.

🚀 Enterprise Blueprint: Native Forms (GLPI 11)

Technical specification for the Service Catalog core integration.

Workflow:

Discovery: AI lists available catalogs using glpi_list_forms.
Setup: AI fetches questions using glpi_get_form_details to build the user interaction.
Submit: AI sends structured answers using glpi_submit_form.

Infrastructure

[ ] Active Notifications: Webhook receiver for push alerts (New critical tickets).
[ ] MySQL Direct Read: Optional engine for high-performance reporting.
[ ] Dockerization: Ready-to-go deployment image.

License

MIT

glpi-mcp-server