Skip to main content

MCP Overview

The slogx MCP server lets AI assistants query your live logs directly — no copy-pasting, no scrolling through terminal output.

The invisible bug problem​

Some bugs only show up in logs. Consider: your app calls an API twice, but the UI deduplicates the response so everything looks fine. The double call is invisible at the UI level — you'd only catch it by noticing duplicate entries in your logs.

With traditional logging, you might never spot this. With slogx + MCP, your AI assistant can search your logs, notice the pattern, and flag it while you're debugging something else entirely.

How it works​

Your AI assistant connects to your running app's slogx WebSocket and queries logs through MCP tools:

You: "Something's wrong with the checkout flow"

Claude: *connects to slogx, searches for checkout-related logs*
Claude: "I see the payment API is being called twice per checkout.
        The second call happens 50ms after the first. Looks like
        a double-trigger from the submit button handler."

The assistant self-serves the information it needs instead of you copy-pasting log output into the chat.

What it provides​

  • connect/disconnect to one or more slogx WebSocket endpoints
  • list active connections and service status
  • token-efficient log search with optional service and level filters
  • fast access to recent errors
  • drill-down into full log details by id

Progressive disclosure for LLMs​

Logs are returned in a compact format (id, timestamp, level, message summary) with full details (args, stacktrace, metadata) available via slogx_get_details. This keeps token usage low while still allowing deep inspection when needed.

Typical flow​

  1. Run your app with slogx live mode enabled.
  2. Start the slogx MCP server.
  3. Add the server to your MCP client config.
  4. Use MCP tools (slogx_search, slogx_get_errors, slogx_get_details) to investigate issues.

Next​