Blowback
Integrates front-end development servers with AI tools, enabling browser automation, console log capture, screenshot management, and HMR event monitoring.
README
Blowback
Vite MCP Server is now Blowback
Blowback aims to support various FE development servers, not only Vite
A Model Context Protocol (MCP) server that integrates FE development servers with AI tools like Claude Desktop and Cursor.
How to Use
Command (Claude Code):
claude mcp add blowback -s project -e PROJECT_ROOT=/path/to/your/project -- npx -y blowback-context
Or use json configuration:
- Claude Code:
{PROJECT_ROOT}/.mcp.json - Cursor:
{PROJECT_ROOT}/.cursor/mcp.json
{
"mcpServers": {
"blowback": {
"command": "npx",
"args": ["-y", "blowback-context"],
"env": {
"PROJECT_ROOT": "/path/to/your/project"
}
}
}
}
Environment Variables
PROJECT_ROOT: Project root path (optional, defaults to current working directory)ENABLE_BASE64: Include base64 encoded images in tool responses (default: false / affects token usage and context window when enabled)
Key Features
- Integration of local development server with MCP server
- Browser console log capture and transmission via MCP
- Checkpoint-based log management
- Screenshot capture and SQLite database management
- HMR (Hot Module Replacement) event monitoring
- Browser automation and element inspection
init Prompt
The init prompt provides guidance to AI assistants on how to effectively use the following features:
Cursor Chat does not support MCP prompt functionality, so this feature is not available. (Claude Code recommended) If needed, manually input the following prompt:
You can use checkpoint features by inserting
<meta name="__mcp_checkpoint" data-id="">into the head to create a named snapshot of the current state. The data-id attribute is a unique identifier for the checkpoint.Console logs generated in the browser while a checkpoint is active are tagged with the checkpoint ID and can be queried individually.
Note: In some development environments, hot reload is triggered when files are saved, so carefully consider the sequence between meta tag changes and the changes you want to observe. Make sure to set the checkpoint meta tag before making the changes you want to track.
You can use the capture-screenshot tool to take screenshots. The captured screenshots are stored in the @.mcp_screenshot/ directory.
Tools
HMR Tools
| Tool Name | Description |
|---|---|
get-hmr-events |
Retrieves recent HMR events |
check-hmr-status |
Checks the HMR status |
Note: HMR connection is optional, not required. HMR event monitoring starts automatically when the browser is launched.
Browser Tools
| Tool Name | Description |
|---|---|
start-browser |
Starts a browser instance and navigates to the development server. HMR monitoring starts automatically |
capture-screenshot |
Captures a screenshot of the current page or a specific element. Returns screenshot ID and resource URI |
get-element-properties |
Retrieves properties and state information of a specific element |
get-element-styles |
Retrieves style information of a specific element |
get-element-dimensions |
Retrieves dimension and position information of a specific element |
monitor-network |
Monitors network requests in the browser for a specified duration |
get-element-html |
Retrieves the HTML content of a specific element and its children |
get-console-logs |
Retrieves console logs from the browser session with optional filtering |
execute-browser-commands |
Safely executes predefined browser commands |
Help Tools
| Tool Name | Description |
|---|---|
how-to-use |
Provides instructions on how to use specific features of the server |
Resources
screenshots
A resource for querying all captured screenshots. You can query screenshot reference IDs captured by the capture-screenshot tool using various criteria.
Images corresponding to reference IDs are managed in the {PROJECT_ROOT}/.mcp_screenshot/ directory.
- URI:
screenshot:// - Returns a list of all screenshots
screenshot-by-url
A resource for querying specific screenshots based on URL path.
Note: Starting from version 1.0, Blob responses through resources are disabled by default, and file reference information is returned instead
- URI template:
screenshot://{+path} - Example:
screenshot://localhost:5173/about - Use URL paths without protocol (http://, https://)
Data Storage Structure
Screenshot Storage
- Screenshot images: Stored in
{PROJECT_ROOT}/.mcp_screenshot/directory - Metadata: Managed in SQLite database in temporary directory
- It's recommended to add
.mcp_screenshot/directory to.gitignore
Log Management System
- Captures browser console logs and saves them to files for querying
- Checkpoint logs are only saved when checkpoints are active
Checkpoint System
How Checkpoints Work
- Checkpoints are used to manage snapshots, logs, screenshots, etc. of specific versions
- When
<meta name="__mcp_checkpoint" data-id="">is inserted into thehead, data is recorded separately using the data-id attribute as an identifier
Architecture and Data Flow
Core Components
-
MCP Server: Central module that exposes tools and resources to AI tools using the Model Context Protocol SDK.
-
Browser Automation: Uses Playwright to control Chrome for visual inspection, screenshot capture, and DOM manipulation.
-
Checkpoint System: Maintains snapshots of browser states for comparison and testing.
-
SQLite Database: Efficiently manages screenshot metadata and enables quick URL-based queries.
Data Sources and State Management
The server maintains several important data stores:
- HMR Event Records: Tracks recent HMR events (updates, errors) from development server.
- Console Message Logs: Captures browser console output for debugging.
- Checkpoint Storage: Stores named snapshots of browser states including DOM snapshots.
- Screenshot Storage: Saves images in project directory and manages metadata with SQLite.
Communication Flow
-
MCP Client → Development Server:
- MCP Client changes the source code and development server detects the change
- Development server automatically updates the browser or emits HMR events
-
Web Browser → MCP Server:
- HMR events and console logs are captured through Playwright
- MCP Server queries the current state of the browser or captures screenshots
-
MCP Server → MCP Client:
- The server converts HMR events into structured responses
- Provides tools for MCP Client to query HMR status, capture screenshots, and more
State Maintenance
The server maintains reference objects for:
- Current browser and page instances
- Recent HMR events
Recommended Servers
playwright-mcp
A Model Context Protocol server that enables LLMs to interact with web pages through structured accessibility snapshots without requiring vision models or screenshots.
Magic Component Platform (MCP)
An AI-powered tool that generates modern UI components from natural language descriptions, integrating with popular IDEs to streamline UI development workflow.
Audiense Insights MCP Server
Enables interaction with Audiense Insights accounts via the Model Context Protocol, facilitating the extraction and analysis of marketing insights and audience data including demographics, behavior, and influencer engagement.
VeyraX MCP
Single MCP tool to connect all your favorite tools: Gmail, Calendar and 40 more.
graphlit-mcp-server
The Model Context Protocol (MCP) Server enables integration between MCP clients and the Graphlit service. Ingest anything from Slack to Gmail to podcast feeds, in addition to web crawling, into a Graphlit project - and then retrieve relevant contents from the MCP client.
Kagi MCP Server
An MCP server that integrates Kagi search capabilities with Claude AI, enabling Claude to perform real-time web searches when answering questions that require up-to-date information.
E2B
Using MCP to run code via e2b.
Neon Database
MCP server for interacting with Neon Management API and databases
Exa Search
A Model Context Protocol (MCP) server lets AI assistants like Claude use the Exa AI Search API for web searches. This setup allows AI models to get real-time web information in a safe and controlled way.
Qdrant Server
This repository is an example of how to create a MCP server for Qdrant, a vector search engine.