[HN Gopher] Introducing Web Search on the Anthropic API
       ___________________________________________________________________
        
       Introducing Web Search on the Anthropic API
        
       Author : cmogni1
       Score  : 112 points
       Date   : 2025-05-07 20:18 UTC (2 hours ago)
        
 (HTM) web link (www.anthropic.com)
 (TXT) w3m dump (www.anthropic.com)
        
       | benjamoon wrote:
       | Good that it has an "allowed domain" list, makes it really
       | useable. The OpenAI Responses api web search doesn't let you
       | limit domains currently so can't make good use of it for client
       | stuff.
        
       | minimaxir wrote:
       | The web search functionality is also available in the backend
       | Workbench (click the wrench Tools icon)
       | https://console.anthropic.com/workbench/
       | 
       | The API request notably includes the exact text it cites from its
       | sources (https://docs.anthropic.com/en/docs/build-with-
       | claude/tool-us...), which is nifty.
       | 
       | Cost-wise it's interesting. $10/1000 queries is much cheaper for
       | heavy use than Google's Gemini (1500 free per day then $35/1000)
       | when you'd expect Google to be the cheaper option.
       | https://ai.google.dev/gemini-api/docs/grounding
        
         | handfuloflight wrote:
         | So the price is just the $0.01 per query? Are they not charging
         | for the tokens loaded into context from the various sources?
        
           | minimaxir wrote:
           | The query cost is in addition to tokens used. It is unclear
           | if the tokens ingested from the search query count as
           | addititional input tokens.
           | 
           | > Web search is available on the Anthropic API for $10 per
           | 1,000 searches, plus standard token costs for search-
           | generated content.
           | 
           | > Each web search counts as one use, regardless of the number
           | of results returned. If an error occurs during web search,
           | the web search will not be billed.
        
         | istjohn wrote:
         | Well also Google has put onerous conditions on their service:
         | 
         | - If you show users text generated by Gemini using Google
         | Search (grounded Gemini), you must display a provided widget
         | with suggested search terms that links directly to Google
         | Search results on google.com.
         | 
         | - You may not modify the text generated by grounded Gemini
         | before displaying it to your users.
         | 
         | - You may not store grounded responses more than 30 days,
         | except for user histories, which can retain responses for up to
         | 6 months.
         | 
         | https://ai.google.dev/gemini-api/terms#grounding-with-google...
         | 
         | https://ai.google.dev/gemini-api/docs/grounding/search-sugge...
        
           | miohtama wrote:
           | Google obviously does not want to cannibalise their golden
           | goose. However it's inevitable that Google search will start
           | to suffer because people need it less and less with LLMs.
        
       | aaronscott wrote:
       | It would be nice if the search provider could be configured. I
       | would like to use this with Kagi.
        
         | lemming wrote:
         | I would really love this too. However I think that the only
         | solution for that is to give it a Kagi search tool, in
         | combination with a web scraping tool, and a loop while it
         | figures out whether it's got the information it needs to answer
         | the question.
        
       | cmogni1 wrote:
       | I think the most interesting thing to me is they have multi-hop
       | search & query refinement built in based on prior
       | context/searches. I'm curious how well this works.
       | 
       | I've built a lot of LLM applications with web browsing in it.
       | Allow/block lists are easy to implement with most web search
       | APIs, but multi-hop gets really hairy (and expensive) to do well
       | because it usually requires context from the URLs themselves.
       | 
       | The thing I'm still not seeing here that makes LLM web browsing
       | particularly difficult is the mismatch between search result
       | relevance vs LLM relevance. Getting a diverse list of links is
       | great when searching Google because there is less context per
       | query, but what I really need from an out-of-the-box LLM web
       | browsing API is reranking based on the richer context provided by
       | a message thread/prompt.
       | 
       | For example, writing an article about the side effects of
       | Accutane should err on the side of pulling in research articles
       | first for higher quality information and not blog posts.
       | 
       | It's possible to do this reranking decently well with LLMs (I do
       | it in my "agents" that I've written), but I haven't seen this
       | highlighted from anyone thus far, including in this announcement.
        
         | simple10 wrote:
         | That's been my experience as well. Web search built into the
         | API is great for convenience, but it would be ideal to be able
         | to provide detailed search and reranking params.
         | 
         | Would be interesting to see comparisons for custom web search
         | RAG vs API. I'm assuming that many of the search "params" of
         | the API could be controlled via prompting?
        
       | simonw wrote:
       | I couldn't see anything in the documentation about whether or not
       | it's allowed to permanently store the results coming back from
       | search.
       | 
       | Presumably this is using Brave under the hood, same as Claude's
       | search feature via the Anthropic apps?
        
         | minimaxir wrote:
         | Given the context/use of encrypted_index and encrypted_context,
         | I suspect search results are temporarily cached.
        
           | simonw wrote:
           | Right, but are there any restrictions on what I can do with
           | them?
           | 
           | Google Gemini has some: https://ai.google.dev/gemini-
           | api/docs/grounding/search-sugge...
           | 
           | OpenAI has some rules too:
           | https://platform.openai.com/docs/guides/tools-web-
           | search#out...
           | 
           | > "When displaying web results or information contained in
           | web results to end users, inline citations must be made
           | clearly visible and clickable in your user interface."
           | 
           | I'm used to search APIs coming with BIG sets of rules on how
           | you can use the results. I'd be surprised but happy if
           | Anthropic didn't have any.
           | 
           | The Brave Search API is a great example of this:
           | https://brave.com/search/api/
           | 
           | They have a special, much more expensive tier called "Data w/
           | storage rights" which is $45 CPM, compared to $5 CPM for the
           | tier that doesn't include those storage rights.
        
             | istjohn wrote:
             | Google's restrictions are outlandish: "[You] will not
             | modify, or intersperse any other content with, the Grounded
             | Results or Search Suggestions..."
        
       | lemming wrote:
       | I'm also interested to know if there are other limitations with
       | this. Gemini, for example, has a built-in web search tool, but it
       | can't be used in combination with other tools, which is a little
       | annoying. o3/o4-mini can't use the search tool at all over the
       | API, which is even more annoying.
        
       | omneity wrote:
       | Related: For those who want to build their own AI search for free
       | and connect it to any model they want, I created a browser MCP
       | that interfaces with major public search engines [0], a SERP MCP
       | if you want, with support for multiple pages of results.
       | 
       | The rate limits of the upstream engines are fine for personal
       | use, and the benefit is it uses the same browser you do, so
       | results are customized to your search habits out-of-the-box (or
       | you could use a blank browser profile).
       | 
       | 0: https://herd.garden/trails/@omneity/serp
        
       | potlee wrote:
       | If you use your own search tool, you would have to pay for input
       | tokens again every time the model decides to search. This would
       | be a big discount if they only charging once for all output as
       | output tokens but seems unclear from the blog post
        
       | jarbus wrote:
       | Is search really that costly to run? $10/1000 searches seems
       | really pricey. I'm wondering if these costs will come down in a
       | few years.
        
         | tuyguntn wrote:
         | they will come down, because up until recently consumers were
         | not paying directly for searches, with the LLM which has a
         | cutoff date in the past and hallucinations, search got popular
         | paid API.
         | 
         | Popularity will grow even more, hence competition will increase
         | and prices will change eventually
        
           | AznHisoka wrote:
           | I dont think that will be true. What competition? Google,
           | Bing, and.. Kagi? (And only one of those have a far superior
           | index/algo than the others)
        
         | jsnell wrote:
         | Yes.
         | 
         | The Bing Search API is priced at $15/1k queries in the cheapest
         | tier, Brave API is $9 at the non-toy tier, Google's pricing for
         | a general search API is unknown but their Search grounding in
         | Gemini costs $35/1k queries.
         | 
         | Search API prices have been going up, not down, over time. The
         | opposite of LLMs, which have gotten 1000x cheaper over the last
         | two years.
        
         | AznHisoka wrote:
         | If you want an unofficial API, most data providers usually
         | charge $4/1000 queries or so. By unofficial, I mean they just
         | scrape whats in Google and return that to you. So thats the
         | benchmark I use, which means the cost here is around 2x that.
         | 
         | As far as I know, the pricing really hasnt gone down over the
         | years. If anything it has gone up because Google is
         | increasingly making it harder for these providers
        
       | zhyder wrote:
       | Now all the big 3 LLM providers provide web search grounding in
       | their APIs, but how do they compare in ranking quality of the
       | retrieved web search results? Anyone run benchmarks here?
       | 
       | Clearly web search ranking is hard after decades of content spam
       | that's been SEO optimized (and we get to look forward to
       | increasing AI spam dominating the web in the future). The best
       | LLM provider in the future could be the one with just the best
       | web search ranking, just like what allowed Google to initially
       | win in search.
        
       ___________________________________________________________________
       (page generated 2025-05-07 23:00 UTC)