Try it
The answer shows up here.
Endpoints
GET/api?q=your question
Answers a question about Shawn, grounded only in what's on this site.
curl "https://groq-chat.shawnpsi.workers.dev/api?q=where+does+shawn+work"
{ "answer": "Shawn is a Data Scientist at C3 AI..." }
GET/api
Everything on the site as one block of text, for your own model to read. No limit.
{ "name": "Shawn Singh", "content": "SHAWN SINGH — data + design...", "usage": "..." }
MCP/mcp
An MCP server with one tool, about_shawn. Add the URL as a connector in Claude, Cursor or any MCP client.
https://groq-chat.shawnpsi.workers.dev/mcp
Importing into ChatGPT Actions, Postman or another tool? Use the OpenAPI spec. Crawlers and agents can start at llms.txt.
CLI
A one-file shell script around the API. Needs only curl. Install:
sudo curl -fsSL https://shawnsingh.me/shawn-ai -o /usr/local/bin/shawn-ai && sudo chmod +x /usr/local/bin/shawn-ai
Then ask away:
shawn-ai "where does shawn work?"
On a Mac with Raycast? Grab the script command:
mkdir -p ~/raycast && curl -fsSL https://shawnsingh.me/ask-shawn.sh -o ~/raycast/ask-shawn.sh && chmod +x ~/raycast/ask-shawn.sh
Add ~/raycast under Raycast Settings → Extensions → Script Commands, then type ask shawn, press Tab and ask.
Rather not install anything? Add &plain to any API call to get bare text back:
curl "https://groq-chat.shawnpsi.workers.dev/api?q=where+does+shawn+work&plain"
Accounts
Signing in to the chat is optional. It saves your conversations so you can pick them up later, and it gives you your own rate limit instead of sharing one with everyone on your network.
How sign-in works. Google and LinkedIn both use OpenID Connect's authorization-code flow, so one code path in the Cloudflare Worker handles both, with no auth library.
- The worker sends you to the provider with a signed
state. The state carries a one-time nonce that is also set as anHttpOnlycookie, and the callback rejects the login unless the two match. That stops a sign-in link started in someone else's browser from signing you into their account. - The worker exchanges the one-time code, plus its client secret, for an ID token directly with the provider over TLS. The token came from the provider itself, so the spec lets the worker skip its signature check. It still checks that the token was issued for this app and hasn't expired.
- Your session is an HMAC-signed token, not a database row, so checking it costs nothing. It comes back in the URL fragment, which browsers never send to servers. It's only ever returned to this site's own pages; any other return address is refused.
- The chat sends the session as an
Authorizationheader rather than a cookie. The site and the API live on different domains, and browsers like Safari block cookies across domains.
What's stored. Your name and email from the provider, plus your saved chats, in a Cloudflare D1 database. Nothing is shared or sold. Delete my data in the chat erases your account and every chat at once. The usage dashboard shows you only your own totals, like answers per day and token counts, never what you asked. The full privacy policy has the details.
Limits
3 questions per minute per IP, 800 characters per question, one question at a time (no chat history). Over the limit you get a 429 with { "error": "..." }. It runs on a free tier, so on busy days it may be rate-limited for everyone. Is it up right now? See the status page.