kagi

name	kagi
description	Kagi API covering privacy-first web search, Universal Summarizer (URL/text/PDF/audio/YouTube, 3 engines, 26 languages), FastGPT (LLM Q&A with cited sources), Web and News Enrichment (non-commercial Teclis/TinyGem indexes), and Small Web RSS feed (free). Triggers on any request for web search, document summarization, AI-answered questions, small-web content, or enrichment of search results. Requires KAGI_API_KEY.
user-invokable	true
argument-hint	kagi search rust async, kagi summarize https://example.com/paper.pdf, kagi ask what is ZTNA, kagi enrich web privacy tools, kagi smallweb
allowed-tools	Bash(curl *)
license	MIT
metadata	{"version":"1.0.1","author":"Brian Morin","homepage":"https://github.com/bdmorin/the-no-shop","category":"search","api_status":"v0 beta","requires":{"env":["KAGI_API_KEY"]},"endpoints":["search","summarizer","fastgpt","enrich-web","enrich-news","smallweb-rss"]}

Kagi API

Base URL: https://kagi.com/api/v0 Small Web RSS base: https://kagi.com/api/v1

API is in v0 beta — breaking changes possible. All responses are JSON.

Authentication

Every request requires:

Authorization: Bot <KAGI_API_KEY>

Token is shown once on generation. Generating a new token invalidates the previous one. Optional IP allowlist configurable in Search API settings.

Get key: https://kagi.com/settings/api Add credits: https://kagi.com/settings/billing_api

Endpoints — Quick Reference

Endpoint	Method	Path	Cost
Search	GET	/search	$0.025/query
Summarizer	GET/POST	/summarize	$0.030/1K tokens (Cecil/Agnes), $1 flat (Muriel)
FastGPT	POST	/fastgpt	$0.015/query
Enrich Web	GET	/enrich/web	$0.002/search (only if results returned)
Enrich News	GET	/enrich/news	$0.002/search (only if results returned)
Small Web RSS	GET	/v1/smallweb/feed/	Free

Universal Response Envelope

Every response follows this structure:

{
  "meta": {
    "id": "uuid",
    "node": "us-east",
    "ms": 143,
    "api_balance": 4.82
  },
  "data": { ... },
  "error": null
}

meta.api_balance — remaining dollars. Track this. error is null on success, an array of error objects on failure.

Error Codes

Code	Meaning
0	Internal error
1	Malformed request
2	Unauthorized
100	No billing information
101	Insufficient credit
200	Summarize failed

Basic Search

curl -H "Authorization: Bot $(printenv KAGI_API_KEY)" \
  "https://kagi.com/api/v0/search?q=your+query&limit=10"

Response data is an array of search objects identified by t field:

• t: 0 — search result (url, title, snippet, published, thumbnail)
• t: 1 — related searches (list array of strings)

Basic Summarize

# Summarize a URL
curl -H "Authorization: Bot $(printenv KAGI_API_KEY)" \
  "https://kagi.com/api/v0/summarize?url=https://example.com/paper.pdf"

# Summarize text (use POST to avoid URL length limits)
curl -X POST -H "Authorization: Bot $(printenv KAGI_API_KEY)" \
  -H "Content-Type: application/json" \
  -d '{"text": "Your long text here...", "engine": "agnes"}' \
  "https://kagi.com/api/v0/summarize"

Basic FastGPT

curl -X POST -H "Authorization: Bot $(printenv KAGI_API_KEY)" \
  -H "Content-Type: application/json" \
  -d '{"query": "What is ZTNA?"}' \
  "https://kagi.com/api/v0/fastgpt"

Image Proxy URLs

API results may return bare image paths — prepend https://kagi.com:

/proxy/filename.jpg?c=HASH  →  https://kagi.com/proxy/filename.jpg?c=HASH

The c= parameter is a cryptographic hash that validates the proxy request.

Key Gotchas

• Search API is invite-only — email [e-mail] for access
• Cached responses are always free — both Summarizer and FastGPT
• Enrichment only billed when results returned — zero-result queries cost nothing
• Summarizer text POST — use POST for text input to avoid URL length limits; max 1MB
• Muriel engine — $1 flat regardless of token count; use for high-stakes summaries
• web_search in FastGPT is currently out of service — web_search: false is the only functioning mode
• Token shown once — store it immediately; can't retrieve it again
• Ultimate plan discount — Summarizer drops to $0.025/1K tokens on Ultimate plan

Reference Files

• references/search-and-enrich.md — Search API params/response, Enrich web/news, Small Web RSS
• references/summarizer.md — Summarizer engines, languages, file types, full params
• references/fastgpt.md — FastGPT params, response schema, reference object