Developer

API v1

Read-only JSON endpoints for prompt search, public collections and automation. No authentication required; requests are limited by IP.

繁體中文

Quick start

Use the curl executable. In Windows PowerShell, use curl.exe if curl is an alias. Quoted URLs keep query characters intact.

Read official model references

curl "https://prompt.luvai.net/api/v1/models?lang=en&limit=1"

Discover tool keys and public counts

curl "https://prompt.luvai.net/api/v1/tools?lang=en&limit=1"

Search published newsletter titles

curl "https://prompt.luvai.net/api/v1/newsletters?lang=en&limit=1"

Search public tutorials

curl "https://prompt.luvai.net/api/v1/tutorials?limit=1"

Search stored engineering terms

curl "https://prompt.luvai.net/api/v1/keywords?q=cache&limit=1"

Find a free prompt

curl "https://prompt.luvai.net/api/v1/prompts?free=1&lang=en&limit=1"

Discover topic keys

curl "https://prompt.luvai.net/api/v1/tags?q=code-review&lang=en"

Browse public collections

curl "https://prompt.luvai.net/api/v1/collections?lang=en&limit=20"

Read one result using its actual slug (Node.js 18+)

Save this code as promptcraft-read.mjs and run node promptcraft-read.mjs. It makes at most two public requests with a ten-second timeout per request. No API key or extra package is needed.

const base = "https://prompt.luvai.net/api/v1";
async function read(url) {
  const response = await fetch(url, { signal: AbortSignal.timeout(10000) });
  if (!response.ok) throw new Error("HTTP " + response.status);
  const data = await response.json();
  if (data.ok !== true) throw new Error("Unexpected API response");
  return data;
}
const page = await read(base + "/prompts?free=1&limit=1&lang=en");
const selected = page.prompts[0];
if (selected) {
  const detail = await read(base + "/prompts/" + encodeURIComponent(selected.slug) + "?lang=en");
  console.log(JSON.stringify(detail, null, 2));
} else {
  console.log("No published free prompts were found.");
}

Paged responses include next_offset. Keep the same filters, use that offset for the next request, and stop when it is null. Collection items are summaries; request /prompts/{slug} for available detail.

Stored engineering glossary

Read public definitions, when-to-use notes and how_to_ask text directly. This performs stored-data lookup, not AI inference, and uses no AI credits. content_language is zh-Hant; English terms are included but definitions remain Traditional Chinese. Returned URLs point to the existing source-language entry.

GET	Query / response
/api/v1/tutorials	q (≤200), limit (1–50; 20), offset; total / showing / next_offset / tutorials
/api/v1/tutorials/{slug}	tutorial: full stored Markdown, content_language=zh-Hant, content_format=markdown, source URL and dates
/api/v1/keywords	q (≤200), category, limit (1–50; 20), offset; total / showing / next_offset / keywords
/api/v1/keywords/{slug}	keyword: definition, when_to_use, how_to_ask, gotcha, related, tags, updated_at

Search is literal: % and _ have no wildcard meaning. category uses the exact values in OpenAPI. Keep filters while following next_offset until null. Search is limited to 30/min/IP and detail to 60/min/IP. Invalid or repeated parameters return 400, missing entries 404, rate limits 429 with Retry-After, and unavailable data 503. No lang parameter or generated translation is provided.

Node.js: discover then read a tutorial

Stored titles, summaries and complete Markdown are Traditional Chinese. Search is literal; no lang option or AI generation. Read through next_offset until null. Returned URLs point to the real source.

const base = "https://prompt.luvai.net/api/v1";
async function read(path) {
  const response = await fetch(base + path, { signal: AbortSignal.timeout(10000) });
  if (!response.ok) throw new Error("HTTP " + response.status);
  const data = await response.json();
  if (data.ok !== true) throw new Error("Unexpected response");
  return data;
}
const page = await read("/tutorials?limit=1");
const selected = page.tutorials[0];
if (selected) {
  const detail = await read("/tutorials/" + encodeURIComponent(selected.slug));
  console.log(JSON.stringify(detail.tutorial, null, 2));
} else {
  console.log("No stored public tutorials were found.");
}

Node.js: search then read a definition

const base = "https://prompt.luvai.net/api/v1";
async function read(path) {
  const response = await fetch(base + path, { signal: AbortSignal.timeout(10000) });
  if (!response.ok) throw new Error("HTTP " + response.status);
  const data = await response.json();
  if (data.ok !== true) throw new Error("Unexpected response");
  return data;
}
const page = await read("/keywords?limit=1");
const selected = page.keywords[0];
if (selected) {
  const detail = await read("/keywords/" + encodeURIComponent(selected.slug));
  console.log(JSON.stringify(detail.keyword, null, 2));
} else {
  console.log("No stored glossary entries were found.");
}

Public newsletter archive

Search stored published issue titles, then read the complete source HTML. lang chooses the source URL interface only; subjects and bodies are unchanged. No subscriber or delivery-operation fields, email sends, subscription changes or AI calls.

GET	Query / response
/api/v1/newsletters	q (≤200 Unicode code points), lang (zh\|en), limit (1–50; 20), offset; total / showing / next_offset / newsletters
/api/v1/newsletters/{slug}	lang (zh\|en); newsletter: subject, published_at, url, content_format=html, content

q matches titles literally (ASCII case-insensitive); % and _ are not wildcards. Keep the same filters and follow next_offset until null. URL-encode each returned slug once in REST paths; use it literally in MCP. Search: 30/min/IP; detail: 60/min/IP. Invalid, unknown or repeated parameters return 400; missing or private issues 404; limits 429 with Retry-After; unavailable data 503. Responses use no-store so publication changes are read again.

Node.js: find then read a published issue

const base = "https://prompt.luvai.net/api/v1";
async function read(path) {
  const response = await fetch(base + path, { signal: AbortSignal.timeout(10000) });
  if (!response.ok) throw new Error("HTTP " + response.status);
  const data = await response.json();
  if (data.ok !== true) throw new Error("Unexpected response");
  return data;
}
const page = await read("/newsletters?limit=1&lang=en");
const selected = page.newsletters[0];
if (selected) {
  const detail = await read("/newsletters/" + encodeURIComponent(selected.slug) + "?lang=en");
  console.log(JSON.stringify(detail.newsletter, null, 2));
} else {
  console.log("No published newsletter issues were found.");
}

Browse the archive MCP: search_newsletters / get_newsletter

Endpoint reference

GET/api/v1/collectionsPublished collection summaries. count is the total before pagination; follow next_offset until it is null.+

Query parameters

Name	Values	Description
lang	zh \| en	Default zh. English fields fall back to the source; translation_status is complete, partial or source.
limit	1-50	Default 20
offset	integer ≥ 0	Default 0; use next_offset from the previous response.