AI Agents Guide

MCP Protocol

Integrate TigerAccess with Model Context Protocol (MCP) for secure LLM agent infrastructure access.

Estimated time: 15 minutes

Key Features

Native MCP server implementation

Tool-level access control

Automatic credential injection

Session context management

Audit logging for all tool calls

Compatible with Claude, GPT, and others

Enable MCP Server

# Start TigerAccess MCP server
tigeraccess mcp start --port 8080

# Or add to config
mcp:
  enabled: true
  port: 8080
  allowed_tools:
    - database_query
    - file_read
    - api_call

Configure Your LLM

# Claude Desktop config (claude_desktop_config.json)
{
  "mcpServers": {
    "tigeraccess": {
      "command": "tigeraccess",
      "args": ["mcp", "start"],
      "env": {
        "TIGERACCESS_API_KEY": "your-api-key"
      }
    }
  }
}

Use Secure Tools

# Your LLM can now use TigerAccess tools
# All calls are authenticated and audited

# Example tool call from LLM:
{
  "tool": "database_query",
  "arguments": {
    "database": "analytics-prod",
    "query": "SELECT COUNT(*) FROM users"
  }
}

MCP Integration Complete

Your LLM agents can now securely access infrastructure through TigerAccess.

View Examples