Every log in one place -
and in your AI's hands.

Quick start

Install in two steps

Install it as a dev dependency:

npx nuxt module add nuxt-spyglass

pnpm add -D nuxt-spyglass

Add it to your Nuxt config - that's the whole setup:

export default defineNuxtConfig({
  modules: ['nuxt-spyglass'],
})

Start your dev server and logs land in .data/spyglass/logs.ndjson (already covered by the default .gitignore).

Configuration

Options

export default defineNuxtConfig({
  modules: ['nuxt-spyglass'],

  spyglass: {
    // Master switch. Always off in production builds, regardless of this value.
    enabled: true,
    // NDJSON log file; relative paths resolve against the project root.
    logFile: '.data/spyglass/logs.ndjson',
    // Rotate once the active file reaches this size (keeps at most two files).
    maxFileSize: 5 * 1024 * 1024,
    // Print the MCP connection state and setup command on startup.
    mcpHint: true,
  },
})

Under the hood

How it works

• Browser: a client plugin wraps console.* and listens for error / unhandledrejection, buffers entries and ships them to a Nitro endpoint.
• Server: a consola reporter plus a Nitro error hook capture server logs and unhandled errors - with their original stack.
• Everything lands in one NDJSON file through a single write queue.
• Correlation: each page load carries a pageLoadId (injected during SSR, sent back to the server on same-origin requests); each request gets its own requestId. Browser and server entries of the same page load therefore share one id.

For humans

Live log viewer (CLI)

A built-in terminal viewer - run it in a second terminal next to your dev server:

npx nuxt-spyglass                  # finds this project's log file by itself
npx nuxt-spyglass <path-to-logs>   # or point it at the file explicitly

• Live tail, newest at the bottom, auto-follows; scroll with ↑/↓, PageUp/PageDown, Home/End, or the mouse wheel.
• Browser + server logs interleaved, colored by level, framework noise dimmed.
• / to search - jump to a hit, which briefly highlights. q or Ctrl+C to quit.

The module prints this command on dev startup as well, next to the log file it writes to.

For your AI

AI access (MCP)

Spyglass ships nuxt-spyglass-mcp, a lightweight stdio MCP server that reads your log file and exposes it to AI agents.

• Your AI agent launches the server itself - you never run it manually.
• Keep your Nuxt dev server running; the MCP server only reads the file on demand.
• After registering, start a new agent session so the tools load.
• No paths, no per-project setup. The server starts in the directory your AI client was opened in and finds that project's log file itself - monorepo layouts like packages/app/.data/spyglass/logs.ndjson included. Register it once, it serves every project you open, always the current one. Pass a path as the first argument (or set SPYGLASS_LOG_FILE) only for a custom logFile.

Tools

Framework noise (Vue warnings, devtools, build lifecycle) is excluded by default; pass includeNoise: true to any tool to see everything.

Set up your agent

claude mcp add --scope user spyglass -- npx -y -p nuxt-spyglass nuxt-spyglass-mcp

Easiest of all: copy the fully-resolved command straight from the dev startup log.

Integration

Works with evlog

Using evlog for structured logging? Spyglass captures it automatically via evlog's evlog:drain hook - reading the structured event directly, so it works even with evlog's console-less default. Per-request logs become individual entries; the wide event becomes a request summary with its fields and error stack, correlated like everything else.

Enable evlog's transport to pull in browser-side evlog logs too:

export default defineNuxtConfig({
  modules: ['nuxt-spyglass', 'evlog/nuxt'],
  // Browser-side evlog logs are captured too once transport is on.
  evlog: { transport: { enabled: true } },
})