Agent Integration

This section helps AI coding agents and developer assistants use the flinkey documentation effectively. It provides structured context, implementation recipes and machine-readable reference material.

How agents can use this documentation

LLM Context

A compact, structured knowledge base that an agent can load into its context window. Covers the integration model, system responsibilities, credential rules and environment variables. Load context →

Implementation Recipes

Step-by-step recipes for authentication, car lifecycle, user lifecycle, assignments, mobile access and Box Commands. Each recipe includes required inputs, expected outputs and agent rules. View recipes →

When to use what

Goal	Resource
Understand the flinkey integration model	LLM Context
Generate backend integration code	Implementation Recipes
Look up exact API operations	API Reference
Understand mobile SDK integration	Mobile SDK
Process webhook events	Webhooks
Use flinkey tools from an AI agent directly	MCP Server
Load full context for external LLMs	/llms.txt

Agent Skill

A ready-to-use Agent Skill is available for download. It enables AI coding agents to build flinkey integrations using the recipes, credential rules and API references from this documentation. Compatible with Cursor, Windsurf, Codex and other agents that support skill files.

Installation:

1. Download the skill files:
   • SKILL.md
   • recipes.md
2. Place them in your project at .cursor/skills/flinkey-integration/ (project-level) or ~/.cursor/skills/flinkey-integration/ (personal)
3. The skill activates automatically when working with flinkey, Tapkey SDK, assignments or Box Commands

Machine-readable files

File	Purpose
/llms.txt	Index file for LLM crawlers — lists key pages and links to the full context
/llms-full.txt	Complete flinkey integration context as a single text file
/openapi.json	OpenAPI 3.1 specification for the flinkey API v3
/keyfob-openapi.yaml	OpenAPI 3.1 specification for the public Keyfob Catalog API
/skills/flinkey-integration/SKILL.md	Agent Skill for flinkey integrations

Agent rules

Danger

AI agents must follow these rules when generating flinkey integration code:

• Never hardcode credentials (API Manager, API-Key, SDK Keys, webhook secrets)
• Never expose backend credentials to mobile apps
• Never log passwords, tokens, API keys or digital keys
• Use placeholders when real values are not available
• Use environment variables for all configuration values
• Do not invent realistic-looking IDs, secrets or tokens

flinkey MCP Server

The flinkey MCP Server exposes all API operations as callable tools — your agent can list cars, create users, manage assignments and query events directly from the IDE.

{
  "mcpServers": {
    "flinkey": {
      "url": "https://flinkey.mcp.witte.digital",
      "headers": {
        "X-Flinkey-API-Key": "{{flinkeyApiKey}}",
        "X-Flinkey-Username": "{{FLINKEY_API_MANAGER_USERNAME}}",
        "X-Flinkey-Password": "{{FLINKEY_API_MANAGER_PASSWORD}}",
        "X-Flinkey-Env": "uat"
      }
    }
  }
}

18 tools available: cars, users, assignments, events, customers and an integration guide.

Full MCP Server documentation →