This portfolio, open to AI agents
The site runs a public MCP (Model Context Protocol) server. Plug in Claude, Cursor, VS Code or any compatible client and ask it directly what I have built, with which technologies, and where.
What it looks like
Three examples of what an agent connected to the site answers. The answers come from the portfolio's real data.
What has Sunny built with React? Give me the links.
sunny · search_portfolio (query: "react")
Seven pieces of work published on his site use React:
- Menu Go! — project, 2020: menus and QR codes for restaurants
- Hey Suzy! — project, 2020: React Native app
- Chimp — project, 2020: Chrome & React
- Recon(vert) — lab, 2020: video converter
- Mes recettes — lab, 2020: voice, recipes and shopping
- Re-Toaster — lab, 2018: toast library
- React ARC — lab, 2017: abstract Redux components
He also works with React at Carrefour, since 2016.
The address
https://sunny.fr/mcpRead-only, no account, no key.
Connect it
Claude Code
claude mcp add --transport http sunny https://sunny.fr/mcpClaude (web, desktop, mobile)
Customize → Connectors → “+” → Add custom connector, then paste the address. No OAuth settings needed.
Other clients
Most clients accept a JSON entry like this one:
{
"mcpServers": {
"sunny": {
"type": "http",
"url": "https://sunny.fr/mcp"
}
}
}Try it here
This panel talks to the real server, live from your browser: the JSON-RPC requests and responses below are exactly what an agent sees.
Tools
Pick a tool on the left, then run it.
Under the hood
- Read-only: the agent reads the portfolio's public content, nothing else. No account, no key.
- No AI on my side: your agent does the thinking, the server just hands it the information.
- Open to WebMCP: browsers that support it can offer the same tools to their agents, right from the site.