A Google search API that hands back the pages, not just the links
Ten links is half an answer
Search results are where research starts, and where most pipelines stall. Someone, or something, still has to open each link, get past the cookie wall and pull out the text. A search API sells you the links, a scraper sells you the pages, and you pay twice to glue the two together.
Search, read and keep, in one place
- Search: a query with its country, language and freshness (the last hour to the last year), and the sites to keep to or leave out.
- Read: ask for the pages and each result is fetched the cheapest way that works, like any MeshArc page, and returned as main-content Markdown.
- Keep: Save and schedule turns a search into a monitor that tells you what is new, what dropped and what moved.
Results you can use straight away
- Organic results as data: position, URL, title, snippet and the engine that answered, for every result.
- News as well as the web: Google’s News tab (Bing News when Google refuses), each result with its publisher and how old it is.
- Pages ready for a model: the article without the navigation, so an agent reads what matters and spends less context.
- A record of every search: listed with its results and what it cost, kept for 7 days.
You pay for answers, not for refusals
| What happens | What it costs |
|---|---|
| Google answers | What its results page took to read — typically 4 credits, a real browser |
| Google refuses and Bing answers | Bing’s results page — typically 1 credit |
| Every engine refuses | Nothing |
| The same search again within the hour | Nothing (within ten minutes for news) |
| Each result’s page, when you ask for it | What that page took to read: 1 credit for a plain fetch, 4 rendered |
Who uses it
- AI and agent builders: ground answers in fresh pages, through the API or the search_web tool in the MeshArc MCP server.
- SEO and growth teams: where a domain ranks for a query, by country, and when it moves.
- Researchers and analysts: the first sources on a question, already as text.
- Sales and partnerships: the companies a query turns up, handed on to the research agent.
Frequently asked questions
Is this a SERP API?
Not in the full sense. It returns the organic results — position, URL, title and snippet — read from the results page. It does not return ads, answer boxes, AI summaries or other results-page features; if you need those, a dedicated SERP API is the better fit.
Which search engines does it use?
Google first. When Google refuses, the same search is read from Bing, and every result says which engine answered.
Can my AI agent use it?
Yes: through the API, or from Claude, Cursor or any MCP client with the search_web tool in the MeshArc MCP server. The MeshArc research agent uses it too, to find pages before it reads them.
How much does a web search cost?
A results page costs what it took to read: typically 4 credits on Google and 1 on Bing. A refused search is free, and an identical search within the hour is answered from the last one at no charge. The pages of the results, when you ask for them, are charged like any page.
Can I track a search over time?
Yes. Save it as a monitor and it runs hourly, daily or weekly, reporting new, dropped and moved results with their positions.
Search, then read, in one call
1,000 free credits, no card. A refused search costs nothing.