One endpoint, POST a message, get a reply. It exists to power the embed widget and it is open to everyone — anonymous by design, rate-limited per IP per day, no signup ever.
POST https://mu7hur.com/api/widget/chat Content-Type: application/json
{ "message": "Explain eigenvectors in one paragraph." }
Response: { "reply": "..." } — plain JSON, CORS enabled, no auth headers.
Try it right now
curl -sS https://mu7hur.com/api/widget/chat -H "Content-Type: application/json" -d "{"message":"Write a haiku about mainframes."}"
That is a real, working AI call. The reply comes from the same family of models that powers the full terminal at mu7hur.com.
Limits, honestly
20 messages / IP / day
Across every site running the widget combined. Enough to try it and build with it lightly; not enough to run a business on. The 21st call returns 429.
2000 chars / message
Messages are truncated to 2000 characters.
Replies up to 4000 chars
One-shot: each message is a fresh conversation, no session state.
Want sessions, streaming, memory, image generation, voice, and the full model lineup? Open the terminal — still no account.
Embed instead of calling
If you want a chat window rather than an API call, the widget is one script tag: