What's different about Metaphor Search?

Metaphor search is fully neural - when you search, we use a transformer-based model to predict the links most relevant to your search. The biggest difference is that in Metaphor, the results you get much more closely match the query you use, letting you for instance search by entity or descriptors.

For example, a Google search for “AI podcasts” returns almost all links like “The 11 Best AI Podcasts” whereas in Metaphor the results are all actual podcasts, neurally sorted by quality and relevance.

How often is the index updated?

We're always crawling the web.

How does Metaphor compare to other search APIs?

Metaphor is fully designed to be used by LLMs:

  • We can instantly return page content for any page in our index.
  • We can return many more results, which LLMs can then sift through.

When should you not use Metaphor?

We think that if you need a web search API, Metaphor is the best option. If you need semantic, neural search, Metaphor search returns much higher quality results. And if you just need traditional search, we wrap traditional search providers with our "keyword" search type and can also get you HTML content from those pages. If you need neither page contents nor neural search, then some other providers may be cheaper.

How can Metaphor be used in an LLM?

Your LLM can make a Metaphor search, see a list of results, and then decide which of those results it wants content for. That content can then be fetched using the /contents endpoint, and your LLM could, for instance, answer a question for a user.

Is it free?

Metaphor is free to use up to 1000 requests per month. Past that, look at our pricing page.

What's our roadmap?

  • Custom comprehensive datasets
  • Support arbitrary non-neural filters.
  • Solve search. No, really.