Transaction Log

Appearance

MCP Integration ​

This document explains how the Txlog Agent implements the Model Context Protocol (MCP) and why this integration is valuable for infrastructure management.

What is MCP? ​

The Model Context Protocol (MCP) is an open standard that enables AI assistants to securely connect with external systems and data sources. Think of it as a "USB-C for AI" – a standardized interface that allows AI models to discover and use tools provided by external applications.

Why MCP for Txlog? ​

Traditional infrastructure queries require:

  1. Logging into a dashboard or terminal
  2. Writing specific queries or commands
  3. Interpreting raw data

With MCP integration, you can simply ask questions in natural language:

  • "Which servers need security updates?"
  • "Show me all changes made to the database servers last week"
  • "Generate a report for the executive team"

The AI assistant translates your intent into the appropriate tool calls and presents the results in a human-readable format.

Architecture ​

text
┌─────────────────┐     ┌─────────────────┐     ┌─────────────────┐
│   AI Assistant  │────▶│   Txlog Agent   │────▶│  Txlog Server   │
│  (Claude, etc)  │     │   (MCP Server)  │     │    (REST API)   │
└─────────────────┘     └─────────────────┘     └─────────────────┘
        │                       │                       │
   Natural Language        Tool Calls              HTTP/JSON
      Queries              + Prompts               Requests

Components ​

  1. AI Assistant: The client application (Claude Desktop, Gemini CLI) that sends natural language queries.

  2. Txlog Agent (MCP Server): Exposes tools and prompts via the MCP protocol. Receives tool invocations and translates them into API calls.

  3. Txlog Server: The data source providing asset information, transaction history, and analytics via REST API.

Transport Modes ​

Stdio Mode ​

Used for local integrations where the AI client spawns the MCP server as a subprocess. Communication happens via standard input/output.

Use case: Claude Desktop integration

text
Claude Desktop ──stdin/stdout──▶ txlog mcp serve

SSE Mode ​

Server-Sent Events mode allows remote clients to connect over HTTP. The MCP server listens on a specified port.

Use case: Web-based AI clients, remote access

text
Web Client ──HTTP/SSE──▶ txlog mcp serve --transport sse --port 3000

Tools vs Prompts ​

Tools ​

Tools are functions that the AI can call to retrieve data or perform actions. They are discoverable and self-describing, allowing the AI to understand what parameters are needed and what data will be returned.

Example: list_assets retrieves a list of servers, optionally filtered by OS.

Prompts ​

Prompts are pre-built templates for complex workflows. They guide the AI through multi-step analysis tasks.

Example: security_audit combines multiple tool calls to produce a comprehensive security assessment.

Security Considerations ​

  1. Authentication: The MCP server uses the same credentials configured in txlog.yaml. Ensure API keys are properly secured.

  2. Data Access: The AI assistant can only access data that the configured user/API key has permission to view.

  3. Read-Only: All MCP tools are read-only operations. The AI cannot modify infrastructure state through this interface.

  4. Local vs Remote: Stdio mode only allows local access. SSE mode should be protected with TLS and access controls when exposed over a network.

Version Requirements ​

The MCP integration requires Txlog Server version 1.19.0 or higher. The agent will check server compatibility on startup and report errors if the server version is insufficient.