Pexels MCP Server
Enables searching and displaying high-quality, royalty-free photos from Pexels directly in ChatGPT conversations via an interactive gallery widget.
README
Pexels ChatGPT App
A ChatGPT App that brings high-quality, royalty-free photography from Pexels directly into your conversations. Built with the Model Context Protocol (MCP) and powered by Cloudflare Workers.

What is This?
This project is a Model Context Protocol (MCP) server that integrates the Pexels API into ChatGPT. Users can search for professional stock photos directly in their chat conversations and view results in an interactive, responsive gallery widget.
Key Features
- š Advanced Photo Search - Search by keywords, orientation, color palette, size, and more
- šØ Interactive Gallery Widget - Beautiful carousel interface with photo cards
- š Theme Aware - Automatically adapts to light/dark mode
- āæ Accessible - Full screen reader support and keyboard navigation
- ā” Serverless - Runs on Cloudflare Workers edge network for global performance
- š± Responsive - Works seamlessly across desktop, tablet, and mobile
Architecture
This project demonstrates a complete MCP application architecture:
āāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāā
ā ChatGPT UI ā
āāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāā¤
ā React Widget (Embedded Gallery) ā
ā ā Uses OpenAI SDK ā
āāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāā¤
ā MCP Protocol (SSE) ā
ā ā Tool Invocation ā
āāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāā¤
ā Cloudflare Worker (Durable Objects) ā
ā ā HTTP Request ā
āāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāā¤
ā Pexels API (https://api.pexels.com) ā
āāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāā
Tech Stack
Backend:
- Cloudflare Workers - Serverless edge computing
- Durable Objects - Stateful MCP server instances
- @modelcontextprotocol/sdk - MCP implementation
- TypeScript 5.9.3
Frontend:
- React 18.3.1 - UI framework
- embla-carousel-react - Carousel functionality
- lucide-react - Icon library
- esbuild - Fast bundler
External API:
- Pexels API - Free stock photography
Project Structure
pexels-app/
āāā src/ # Backend (Cloudflare Worker)
ā āāā index.ts # MCP Durable Object & worker entry
ā āāā types.ts # Shared TypeScript types
ā āāā tools/
ā ā āāā pexels.ts # Pexels search tool implementation
ā āāā components/
ā āāā react-widget-inline.ts # Bundled React widget (auto-generated)
ā āāā react-widget-resource.ts # HTML wrapper for widget
ā
āāā web/ # Frontend (React Widget)
ā āāā src/
ā ā āāā component.tsx # React entry point
ā ā āāā theme.tsx # Theme tokens & context
ā ā āāā components/
ā ā ā āāā App.tsx # Main gallery application
ā ā ā āāā cards/ # Photo card components
ā ā āāā hooks/
ā ā āāā use-openai-global.ts # OpenAI SDK integration
ā ā āāā use-widget-state.ts # Persistent state management
ā āāā dist/ # Build output
ā
āāā scripts/
ā āāā inline-react-widget.js # Bundles React into Worker
ā
āāā docs/ # Comprehensive documentation
ā āāā ARCHITECTURE.md # System design
ā āāā DEPLOYMENT-GUIDE.md # Deployment instructions
ā āāā ... # More guides
ā
āāā wrangler.jsonc # Cloudflare Worker config
āāā .dev.vars.example # Environment variable template
āāā package.json # Dependencies & scripts
Getting Started
Prerequisites
- Node.js 18+ and npm
- Pexels API key (free)
- Cloudflare account (free tier available)
Installation
-
Clone the repository:
git clone https://github.com/yourusername/pexels-app.git cd pexels-app -
Install dependencies:
npm install cd web && npm install && cd .. -
Configure environment variables:
cp .dev.vars.example .dev.varsEdit
.dev.varsand add your Pexels API key:PEXELS_API_KEY=your_actual_api_key_here -
Build and run locally:
npm run devThe MCP server will be available at
http://localhost:8787
Deployment
-
Authenticate with Cloudflare:
npx wrangler login -
Set production secrets:
npx wrangler secret put PEXELS_API_KEY # Enter your API key when prompted -
Deploy to Cloudflare:
npm run deployYour app will be live at
https://your-worker-name.workers.dev
Usage
MCP Tool: pexels.searchPhotos
Search the Pexels library with powerful filtering options:
Parameters:
| Parameter | Type | Required | Description |
|---|---|---|---|
query |
string | Yes | Search keywords (1-120 characters) |
page |
number | No | Page number (1-50, default: 1) |
perPage |
number | No | Results per page (1-30, default: 12) |
orientation |
string | No | landscape, portrait, or square |
size |
string | No | large, medium, or small |
color |
string | No | Hex code or color keyword |
locale |
string | No | ISO locale code (e.g., en-US) |
Example in ChatGPT:
User: Show me portrait photos of mountain landscapes
[ChatGPT invokes: pexels.searchPhotos({
query: "mountain landscapes",
orientation: "portrait",
perPage: 15
})]
[Interactive gallery widget displays with 15 photos]
Output:
- Text summary: "Found X Pexels photos for 'query'"
- Interactive gallery widget with photo cards
- Each card shows:
- High-quality photo preview
- Photographer attribution
- "View on Pexels" button (opens in new tab)
Development
NPM Scripts
# Development
npm start # Start local dev server
npm run dev # Same as start
# Building
npm run build:widget # Build React widget only
npm run deploy # Build everything and deploy
# React Development (in web/ directory)
cd web
npm run build # Build React bundle
npm run watch # Watch mode for React changes
# Type Checking
npm run type-check # Check TypeScript types
Development Workflow
- Make changes to React widget (
web/src/) - Build widget:
cd web && npm run build - Inline bundle:
node scripts/inline-react-widget.js - Test locally:
npm run dev - Deploy:
npm run deploy
For faster iteration, use watch mode in a separate terminal:
cd web && npm run watch
Configuration
Environment Variables
Development (.dev.vars):
PEXELS_API_KEY=your_api_key_here
Production (Cloudflare Secrets):
npx wrangler secret put PEXELS_API_KEY
npx wrangler secret list # View configured secrets
Wrangler Configuration
Key settings in wrangler.jsonc:
{
"name": "pexels",
"main": "src/index.ts",
"compatibility_date": "2025-03-10",
"durable_objects": {
"bindings": [{
"class_name": "MyMCP",
"name": "MCP_OBJECT"
}]
},
"vars": {
"PEXELS_API_BASE_URL": "https://api.pexels.com/v1"
}
}
Documentation
Comprehensive documentation is available in the docs/ directory:
- ARCHITECTURE.md - How frontend and backend connect
- DEPLOYMENT-GUIDE.md - Complete deployment instructions
- PROJECT-TOUR.md - File-by-file walkthrough
- APP_DESIGN_GUIDELINES.md - OpenAI design principles
- METADATA-OPTIMIZATION.md - MCP metadata tuning
- FUTURE-UI-UX-GUIDE.md - UI styling best practices
- WRANGLER-SECRET-COMMANDS.md - Secret management
Design Philosophy
This project follows OpenAI's ChatGPT App design guidelines:
- Conversational - Seamlessly integrated into chat flow
- Intelligent - Context-aware tool invocation
- Simple - Focused, single-purpose interactions
- Responsive - Fast, lightweight, edge-optimized
- Accessible - Screen reader support, keyboard navigation
UI/UX Principles
- System-First Design - Inherits ChatGPT's colors, fonts, and spacing
- Theme Tokens - Dynamic light/dark theme support
- Transparent Backgrounds - Blends naturally with host environment
- Component-Scoped Styles - No global CSS to avoid conflicts
- Minimal Branding - Only subtle accent colors on CTAs
API Reference
Pexels API Integration
The tool makes requests to the Pexels API v1:
Endpoint: GET https://api.pexels.com/v1/search
Authentication: Bearer token via Authorization header
Rate Limits: 200 requests/hour (free tier)
Response Format: Normalized from snake_case to camelCase for TypeScript
For full Pexels API documentation, visit: https://www.pexels.com/api/documentation/
Advanced Features
Custom React Hooks
useOpenAiGlobal(key) - Access OpenAI SDK globals
const theme = useOpenAiGlobal('theme'); // 'light' | 'dark'
const toolOutput = useOpenAiGlobal('toolOutput'); // Tool result data
useWidgetState(defaultState) - Persistent widget state
const [state, setState] = useWidgetState({ page: 1 });
// State survives widget remounts
useThemeTokens() - Access theme design tokens
const tokens = useThemeTokens();
// tokens.colors.background, tokens.fonts.body, etc.
Type Safety
- Full TypeScript coverage with strict mode
- Zod validation for all tool inputs
- Shared types between frontend and backend
- Automatic type inference from OpenAI SDK
Error Handling
Graceful handling of:
- API authentication failures
- Network errors
- Missing configuration
- Invalid parameters
- Empty search results
- User-friendly error messages in widget
Security
Content Security Policy
The widget enforces a strict CSP:
default-src 'none';
script-src 'self' 'unsafe-inline';
style-src 'self' 'unsafe-inline';
img-src 'self' data: https://images.pexels.com;
frame-ancestors 'none';
Best Practices
- API keys stored as Cloudflare secrets (never in code)
.dev.vars.exampleprovided as template- Input validation with Zod schemas
License
This project is licensed under the MIT License - see the LICENSE file for details.
Acknowledgments
- Pexels - For providing free, high-quality stock photography
- Cloudflare Workers - For serverless infrastructure
- OpenAI - For ChatGPT and the MCP specification
- Model Context Protocol - For the protocol specification
Learn More
MCP Resources
Cloudflare Resources
ChatGPT App Development
Built with ā¤ļø using the Model Context Protocol
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.