Google search API built for AI agents
SerpKite is a search API built for AI agents. One key gives you Google web, news, images, videos, maps, places, reviews, shopping, scholar and patents, plus webpage-to-Markdown extraction, site maps, async crawls and scheduled monitors. Results come back as full JSON, token-lean compact JSON or Markdown, so agents spend fewer tokens. Prepaid credits never expire, and failed or empty searches are never billed.
12 Google verticals under one /v1 API. LLM-ready output: JSON, compact JSON or Markdown. include_content returns top results as clean Markdown in the same call. Extract, map and crawl websites; scheduled monitors with webhooks. Remote MCP server (14 tools) for Claude, Cursor and VS Code. SDKs for TypeScript, Python and Go plus LangChain, CrewAI, Vercel AI SDK, n8n and Dify integrations.
Give AI agents and RAG pipelines fresh Google results. Ground chatbots in current web data. Track keyword rankings and news mentions with scheduled monitors. Turn any webpage or whole site into clean Markdown for LLMs. Add web search to Claude, Cursor or VS Code through MCP.
A launchpad for independent makers and startups to publish p
AI chatbots for customer messaging. No subscription fees.
Know the next move to get your first visitors.
Comments
1Hi Fazier! I'm building SerpKite because every agent I built needed fresh Google results, and the existing SERP APIs returned huge HTML-shaped JSON that burned tokens, billed failed searches and expired unused credits. SerpKite returns compact JSON or Markdown, bills only successful results, and the credits never expire. There's a remote MCP server so Claude/Cursor can search out of the box. 1,000 free credits on signup, tell us what breaks.