Debug with AI

Turn your live NgRx DevTool session into AI-ready debugging context with no API key, subscription, or extra install required.

Copy a debugging prompt

Click Debug with AI in the toolbar. The DevTool assembles a focused prompt from the current session:

• Recent actions
• Effect errors with stack traces
• Slowest renders
• Latest state diff

Paste the prompt into GitHub Copilot Chat, Claude, ChatGPT, or another assistant. Pick a focus such as health check, debug errors, why slow, or explain flow, and the prompt adapts.

Run locally with Ollama

If you have Ollama running, the dialog detects it automatically and lets you ask a local model right inside the DevTool. Answers stream in, and nothing leaves your machine.

ollama pull llama3

Connect via MCP

The server exposes a Model Context Protocol endpoint at http://localhost:3000/mcp, so AI assistants can query your running NgRx session directly.

Add this to .vscode/mcp.json in your project:

{
  "servers": {
    "ngrx-devtool": {
      "type": "http",
      "url": "http://localhost:3000/mcp"
    }
  }
}

Then ask Copilot Chat things like:

Why is my loadBooks$ effect failing?

MCP tools

Tool	What it returns
get_debug_summary	Recent actions, effect errors and performance at a glance
get_action_timeline	Recent actions with source, render time and warnings
get_effects / get_effect_errors	Effect executions, plus errors with stack traces
get_state_diff	What changed in the store for a given action
get_current_state	The latest store snapshot
get_performance_stats	Average/max render time and the slowest actions

info

The MCP server captures traffic in memory on the DevTool server and adds no new dependencies to the published package.