For AI agents
How to navigate this site: machine-readable formats and endpoints are listed below.
Markdown
Markdown for this page: https://harpie.us/deploy/index.html.md. Other pages use the same pattern: append .md to the path (home: /index.md).
Site briefing (llms.txt)
Site-wide briefing: https://harpie.us/llms.txt (also /agents.md) — product summary, key pages, and brand guidelines.
Ask
Product questions: https://harpie.us/ask. Send one question; the reply is Markdown grounded in published product facts.
How to call it:
- Question only:
POST /ask with Content-Type: text/plain and the question as the body
- Question with context:
POST /ask with Content-Type: application/json. Send query (required question), context (optional caller-provided context), and url (optional page URL on this website).
- Use
context for the user's requirements, preferences, or a relevant text excerpt to tailor the answer. Limits: query 1,000 characters, context 8,000 characters, url 2,048 characters.
- JSON example:
{"query": "Which plan fits this team?", "context": "25 employees; SSO required."}
- Compatibility:
GET /ask?q=<URL-encoded-question> for non-sensitive questions up to 1,000 characters
- Reply: raw Markdown grounded in this brand's current product facts
Machine-readable protocols
- Site briefing: /llms.txt (alias /agents.md)
- WebMCP (in-browser agents): /murmur-agent.js
- Ask directly: /ask
- A2A: Read /.well-known/agent-card.json and use its advertised transport and URL. Send A2A-Version: 1.0. The legacy discovery URL /.well-known/agent.json also works.
For product-fit or buying questions about Harpie, POST to /ask.