Submit your tool

Daily brief

Best web search tool for Claude and OpenAI agents, November 2026: prices and limits of 13 APIs checked

On November 12, 2026, Claude and OpenAI both list web search at $10 per 1,000 calls, before the tokens those results become. The cheapest printed result-list price in this set is $1 per 1,000.

Checked November 12, 2026. Prices, free allowances, and rate limits were read that day from each vendor’s own pricing or docs page. GitHub stars and commit times were read from the repository API the same day. Nothing here is estimated.

Published Updated 8 min readAI Agents Info Research

  • $10

    Claude and OpenAI, per 1,000 searches

    Tool fee. Retrieved text bills as tokens. OpenAI’s non-reasoning preview is $25.

  • $1

    Cheapest printed result list, per 1,000

    Parallel fast or turbo, and Perplexity Search with search_type fast.

  • $14

    Gemini 3 and newer grounding, per 1,000

    After 5,000 free requests a month. Daily cap 1,500. Each query can bill.

  • 7,803

    Stars on Firecrawl’s official MCP server

    Default-branch commit 3.40.0 at 03:19 UTC on 12 Nov 2026.

Executive summary

On 12 November 2026 this page records one web-search call across thirteen APIs: Claude, OpenAI, Gemini, Perplexity, Tavily, Exa, Brave, Parallel, Firecrawl, SerpAPI, Linkup, You.com, and Jina. 135

Claude and OpenAI charge $10 per 1,000, then tokens. Parallel fast and Perplexity fast print $1 per 1,000. Gemini 3 and newer prints $14 per 1,000 after 5,000 free a month. 1717

Nothing here is estimated. Printed prices are copied. A figure this page multiplies or divides is labeled computed. A missing dollar rate is marked not published.

Key findings

  1. 01Anthropic logoAnthropic lists web search at $10 USD per 1,000 searches, plus tokens for retrieved text. OpenAI logoOpenAI lists the same $10.00 per 1,000 calls. OpenAI logoOpenAI’s preview on non-reasoning models is $25.00 per 1,000, and those search-content tokens are free.
  2. 02Parallel lists turbo and fast at $1 per 1,000 requests, and basic and advanced at $5. Perplexity logoPerplexity Search is $5.00 per 1,000, or $1.00 on fast. Up to five queries in one request are one bill.
  3. 03Gemini logoGemini 3 and newer grounding is 5,000 free requests a month, then $14 per 1,000, capped at 1,500 a day. One prompt can run several queries. Gemini logoGemini 2.5 is $35 per 1,000 prompts after 1,500 free a day.
  4. 04firecrawl/firecrawl-mcp-server had 7,803 GitHub logoGitHub stars. Its newest commit on main is 3.40.0, at 03:19 UTC on 12 November 2026. The hosted Claude logoClaude, OpenAI logoOpenAI, and Gemini logoGemini tools had no MCP repo in an org search that day.
  5. 05Jina’s public pricing page prints a floor of 10,000 tokens per s.jina.ai request, and 100 requests a minute with a free key. The dollar rate is not published.

What does one search call cost on November 12?

Each bar is one mode, before tokens and extra results. SerpAPI’s bar is Starter, $25 for 1,000 searches. Jina’s dollar rate is not published.

Dollars per 1,000 calls. A longer bar is a higher price. Tavily’s $8 is 1,000 × $0.008, computed. Firecrawl’s plan arithmetic is in the table.
List price per 1,000 search callsDollars per 1,000 calls. A longer bar is a higher price. Tavily’s $8 is 1,000 × $0.008, computed. Firecrawl’s plan arithmetic is in the table. Unit: USD per 1,000 calls. Vendor pricing pages. Checked 12 Nov 2026. Tavily: 1,000 × $0.008, computed. [[1]][[3]][[5]][[7]][[12]][[14]][[16]][[17]][[20]][[22]][[24]]Parallel fast$1Perplexity fast$1Exa lite$4Brave Search$5Linkup standard$5You.com search$5Tavily basic$8OpenAI web_search$10Claude web_search$10Gemini 3+ search$14SerpAPI Starter$25
List price per 1,000 search calls
CategoryUSD per 1,000
Parallel fast$1
Perplexity fast$1
Exa lite$4
Brave Search$5
Linkup standard$5
You.com search$5
Tavily basic$8
OpenAI web_search$10
Claude web_search$10
Gemini 3+ search$14
SerpAPI Starter$25

Vendor pricing pages. Checked 12 Nov 2026. Tavily: 1,000 × $0.008, computed. 135712141617202224 Unit: USD per 1,000 calls.

Tavily pay as you go is $0.008 a credit. Basic search is 1 credit, so $8 per 1,000, computed. Project prints $0.0075, or $7.50 per 1,000, computed. Advanced is 2 credits. 12

Firecrawl search is 2 credits per 10 results. Standard, $99 for 50,000 searches, is $1.98 per 1,000 if spent on search, computed. Hobby is $19 for 2,500 searches, $7.60 per 1,000, computed. A gov-category search adds no search credits. 1819

List price of one search call, read 12 November 2026. Computed cells say so. “not published” means the official page did not print the figure.
APIPrice per 1,000 callsFree startWhat comes backSourceChecked
Claude web_search$10, plus tokens for retrieved textnot publishedCited results. Citation fields are outside the token count. A failed search is not billed.Anthropic pricing 3412 Nov 2026
OpenAI web_search$10, plus search-content tokens. Preview on non-reasoning models: $25, content tokens free.not publishedA web_search_call, then a message with url citations. Search context window is 128,000 tokens.OpenAI pricing 1212 Nov 2026
Gemini grounding, Gemini 3 and newer$14 after 5,000 free requests a month5,000 requests a month on the paid tier. Free tier: not available.A grounded answer. Each search query the model runs can bill separately.Gemini pricing 5612 Nov 2026
Perplexity Search API$5 standard, or $1 with search_type fastnot publishedRanked JSON with snippet text. Up to five queries are one bill. No model tokens.Perplexity pricing 712 Nov 2026
Tavily Search, basic$8 on pay as you go, computed1,000 credits a month, no cardBasic is 1 credit. Advanced is 2. The credits page does not describe the JSON.Tavily credits 1212 Nov 2026
Exa /search, lite$4 for up to 10 results. Auto and fast are $7.$10 in credits a month, up to 2,500 Lite searches, no cardSearch results with page contents. Results past 10 are $1 per 1,000 results.Exa pricing 1412 Nov 2026
Brave Search$5not publishedURLs, text, news, images, and LLM context. Answers are $4 per 1,000 plus tokens.Brave plans 1612 Nov 2026
Parallel Search, fast or turbo$1 for 10 results. Basic and advanced are $5.not publishedPages and excerpts. Results past 10 are $1 per 1,000. Excerpt length does not change the price.Parallel pricing 1712 Nov 2026
Firecrawl /search2 credits per 10 results. Standard computes to $1.98 if the plan is used on search.1,000 credits a month, no card, listed as 500 searchesTitle, URL, and description, with highlights by default. Markdown is a scrape.Firecrawl pricing 181912 Nov 2026
SerpAPIStarter is $25 for 1,000 searches. Production is $10 per 1,000, computed, if all 15,000 are used.250 searches a monthOne credit per successful search, whether the response has 100 results or none.SerpAPI pricing 2012 Nov 2026
Linkup /search, standard$5 for searchResults. Sourced answer or structured output is $6. Deep searchResults is $50.$20 for a professional email, topped up to $20 each monthRanked sources, or a sourced answer, or JSON. The FAQ prints the per-1,000 figures.Linkup pricing and FAQ 212212 Nov 2026
You.com Web Search$5, up to 100 results. Live full-page extraction is $1 per 1,000 pages.100 queries a day through the MCP server, no key and no cardWeb and news JSON. Snippets by default. One result and 100 results are the same call price.You.com billing and pricing 232412 Nov 2026
Jina s.jina.ainot publishedA free key is offered. The token grant’s dollar value is not printed.LLM-friendly text of search results. Each request starts at 10,000 tokens.Jina pricing 2512 Nov 2026

What do the Claude and OpenAI tools add to the call?

New integrations send type web_search. The preview tool skips filters and return_token_budget. On gpt-4o-mini and gpt-4.1-mini, non-preview search content is 8,000 input tokens a call. The search window is 128,000 tokens. Limits follow the model tier. 12

Each Claude search is one use. Errors are not billed. Later turns must return encrypted_content unchanged. cited_text, title, and url sit outside the token count. Retrieved text is input tokens. 34

Gemini 3 bills each query the model runs. Empty queries are ignored. Gemini 2.5 bills per prompt, $35 per 1,000 after 1,500 free a day. Search Grounding excludes retrieved tokens. Claude and OpenAI bill that text. 56

perplexity/sonar is $0.25 and $2.50 per million tokens. Chat Completions support ended 27 September 2026. Agent web_search is $2.50 per 1,000, or $1.00 on fast, plus tokens. The raw Search API has no token line. 7910

Which limit stops an agent loop first?

A cheap call still stalls when the public ceiling is low. A silent page is marked not published.

Published rate limits, read 12 November 2026. A blank public page is “not published,” not a guess.
APIPublished limitSourceChecked
OpenAI web_searchThe underlying model’s tiered rate limits. No separate search RPM on the guide.OpenAI guide 212 Nov 2026
Claude web_searchThe organization’s console limit. No single public RPM.Claude tool page 412 Nov 2026
Gemini 3 and newer grounding1,500 requests a day per project, after the pricing row’s monthly allowance.Gemini pricing 512 Nov 2026
Perplexity Search API50 query units a second, every tier. Five queries spend five units and one bill.Perplexity rate limits 812 Nov 2026
Perplexity Agent API1 QPS at Tier 0 ($0 purchased), up to 33 QPS at Tier 4 and Tier 5.Perplexity rate limits 812 Nov 2026
Tavily100 RPM on the free Researcher plan. 1,000 RPM once pay as you go is on.Tavily rate limits 1312 Nov 2026
Exa /search10 QPS. Deep types are 5 QPS. 25 QPS after $1,000 purchased in 30 days.Exa billing 1512 Nov 2026
Brave Search50 queries a second. The Answers endpoint is 2 queries a second.Brave plans 1612 Nov 2026
Parallel Searchnot publishedParallel pricing 1712 Nov 2026
Firecrawl /search10 a minute on Free, 100 on Hobby, 500 on Standard. Concurrency is 2 on Free.Firecrawl pricing 1812 Nov 2026
SerpAPI50 successful searches an hour on Free, 200 on Starter, 3,000 on Production.SerpAPI pricing 2012 Nov 2026
Linkupnot publishedLinkup pricing 2112 Nov 2026
You.com Web Search100 queries a day on the free MCP profile. Paid QPS is a custom enterprise limit.You.com pricing 2412 Nov 2026
Jina s.jina.aiNo key is blocked. 100 RPM on a free or paid key. 1,000 RPM on premium.Jina pricing 2512 Nov 2026

Gemini caps grounding at 1,500 requests a day. Four queries a task fill that day in 375 tasks. Perplexity Search allows 50 query units a second on every tier. 58

Which vendors publish an official MCP server?

Stars are stargazers_count. The time is the newest commit on main. pushed_at is newer for Exa (5 Nov vs 3 Nov), SerpAPI (10 Nov vs 4 Nov), and Linkup (1 Nov vs 1 Oct). The table uses the commit. 26

Official MCP repositories named by the vendor org or the repo description, read from the GitHub API on 12 November 2026.
RepositoryStarsNewest commit on main (UTC)SourceChecked
firecrawl/firecrawl-mcp-server7,8032026-11-12 03:19GitHub API 262712 Nov 2026
exa-labs/exa-mcp-server5,2272026-11-03 01:15GitHub API 2612 Nov 2026
perplexityai/modelcontextprotocol2,6212026-10-21 23:47GitHub API 2612 Nov 2026
tavily-ai/tavily-mcp2,4862026-11-05 16:36GitHub API 262812 Nov 2026
brave/brave-search-mcp-server1,5692026-11-09 12:07GitHub API 2612 Nov 2026
jina-ai/MCP9072026-10-15 22:19GitHub API 2612 Nov 2026
serpapi/serpapi-mcp1842026-11-04 10:04GitHub API 2612 Nov 2026
parallel-web/search-mcp322026-10-05 23:15GitHub API 2612 Nov 2026
LinkupPlatform/linkup-mcp-server292026-10-01 14:17GitHub API 2612 Nov 2026
youdotcom-oss/mcp272026-10-28 16:20GitHub API 2612 Nov 2026

Hosted servers: Exa at https://mcp.exa.ai/mcp and You.com at https://api.you.com/mcp?profile=free. Parallel’s repo says “Free web search and page fetching for agents. Hosted MCP, no API key required.” The Search API price page leaves that allowance not published. 141723

On 12 November 2026, GitHub search returned 0 repos for the OpenAI, Anthropic, Google, and Google APIs org queries. Those hosted tools have no official MCP repo in that search. 26

What changed from November 5 through November 12?

No pricing page re-read today carries a price change dated 5–12 November 2026. OpenAI’s entry in that window is GPT-6-Cyber on 5 November. You.com’s newest changelog line is 26 October 2026. 223

Perplexity’s November changelog names ultra@2026-11-05.0 on claude-opus-5-5, and 2026-11-11.0 versions that can omit max_output_tokens. gemini-4-carbon is $2, $0.10 cached, and $10 per million tokens through 30 November 2026. Search stays $5 and $1. 11

The new 2026-11-11.0 versions of fast, low, medium, high, xhigh, and ultra use this default. Calling an effort level by name, such as effort="fast", selects the new version.

Perplexity API changelog, November 2026. Checked 12 Nov 2026. 11

Firecrawl MCP 3.40.0 landed at 03:19 UTC on 12 November 2026: Alexandria tools, max_credits guidance, and people and company search. Ordinary search stays 2 credits per 10 results. Alexandria discovery is free. 1927

Tavily’s MCP repo merged an OAuth docs change at 16:36 UTC on 5 November 2026. Brave’s 9 November commit bumps a dependency. SerpAPI v2.2.0 is dated 4 November, outside this window. 2628

What was left out?

No query was sent and no account was billed. Quality scores, including Linkup’s FAQ benchmarks, stay off this page. Serper was not re-fetched. Consumer chat plans stay on the buyer’s comparison.

Not published: Jina’s dollar rate, Brave’s free credit, Parallel’s free allowance and rate limit, Linkup’s rate limit, You.com’s paid QPS, and a free search allowance for OpenAI or Claude.

Firecrawl’s $1.98 and SerpAPI Production’s $10 assume the monthly allowance is used. Tavily’s $8 is 1,000 times the printed credit price.

FAQ

What is the cheapest web search tool for a Claude or OpenAI agent?

Parallel fast or turbo, and Perplexity Search with fast, print $1 per 1,000. Claude and OpenAI print $10 per 1,000, plus tokens.

Do Claude and OpenAI charge the same amount for web search?

Both tool fees are $10 per 1,000. OpenAI’s non-reasoning preview is $25 per 1,000, with content tokens free. Mini models bill 8,000 input tokens a call.

How much is Gemini grounding with Google Search?

Gemini 3 and newer: 5,000 free a month, then $14 per 1,000, capped at 1,500 a day. Gemini 2.5 is $35 per 1,000 prompts after 1,500 free a day.

Is Perplexity Sonar still a separate search price?

Chat Completions support ended 27 September 2026. perplexity/sonar remains $0.25 and $2.50 per million tokens. Agent web_search is $2.50 or $1.00 per 1,000. The raw Search API adds no model tokens.

Which of these publish an official MCP server?

Ten vendor orgs had a search MCP repo on 12 November 2026. Firecrawl’s had 7,803 stars. The hosted Claude, OpenAI, and Gemini tools had none.

Methodology

Every dollar figure was read on 12 November 2026 from the vendor’s own page. A multiplied or divided figure is labeled computed. A missing rate is marked not published.

GitHub stars are stargazers_count. Commit time is the newest commit on the default branch. The window is 5–12 November 2026. Nothing here is estimated. Computed: Tavily $8 and $7.50, Firecrawl $1.98 and $7.60, SerpAPI Production $10.

Sources

  1. 1OpenAI, API pricing. Checked 12 November 2026. Web search $10.00 per 1,000 calls plus search-content tokens. Preview on reasoning models $10.00. Preview on other models $25.00, search-content tokens free. gpt-4o-mini and gpt-4.1-mini non-preview search content is a fixed 8,000 input tokens.
  2. 2OpenAI, Web search guide. Checked 12 November 2026. New integrations use type web_search. Search context window 128,000 tokens. Rate limits match the model tier. Changelog in the same docs: 5 November is GPT-6-Cyber, not a search-price change.
  3. 3Anthropic, API pricing. Checked 12 November 2026. Web search is $10 USD per 1,000 searches, plus token costs. Each search is one use. A failed search is not billed.
  4. 4Anthropic, Web search tool. Checked 12 November 2026. max_uses, encrypted_content, and the rule that cited_text, title, and url do not count as tokens. The organization rate limit is on the console page.
  5. 5Google, Gemini API pricing. Checked 12 November 2026. Gemini 3 and newer: 5,000 free search requests a month, then $14 per 1,000, daily cap 1,500 per project. Gemini 2.5: 1,500 RPD free, then $35 per 1,000 grounded prompts. Agents row: Search Grounding retrieved tokens excluded.
  6. 6Google, Grounding with Google Search. Checked 12 November 2026. Gemini 3 bills each search query the model executes. Gemini 2.5 and older models bill per prompt.
  7. 7Perplexity, Pricing. Checked 12 November 2026. Search API $5.00 per 1,000 requests, $1.00 with fast. Up to five queries are one bill. Agent web_search $0.0025, or $0.001 on fast.
  8. 8Perplexity, Rate limits. Checked 12 November 2026. Search API: 50 query units a second, all tiers. Agent API: 1 QPS at Tier 0 through 33 QPS at Tier 4 and Tier 5. A five-query request spends five rate-limit units.
  9. 9Perplexity, Agent API models. Checked 12 November 2026. perplexity/sonar: $0.25 per million input tokens, $2.50 per million output tokens, $0.0625 per million cache-read tokens.
  10. 10Perplexity, Migrate from Sonar. Checked 12 November 2026. Sonar Chat Completions support ended 27 September 2026. Synchronous requests are reformulated as Agent API requests.
  11. 11Perplexity, Changelog. Checked 12 November 2026. November 2026 entries: ultra@2026-11-05.0, effort versions 2026-11-11.0, and google/gemini-4-carbon introductory token prices through 30 November 2026.
  12. 12Tavily, Credits and pricing. Checked 12 November 2026. Free 1,000 credits a month, no card. Pay as you go $0.008 a credit. Basic search 1 credit. Advanced search 2 credits. Project plan $30 for 4,000 credits at $0.0075 a credit.
  13. 13Tavily, Rate limits. Checked 12 November 2026. Researcher plan 100 requests a minute. Pay as you go 1,000. Growth and Scale 1,500.
  14. 14Exa, Pricing. Checked 12 November 2026. Lite and instant $4 per 1,000 requests for up to 10 results. Auto and fast $7. Deep and deep-lite $12. Deep-reasoning $15. Free tier $10 a month, up to 2,500 Lite searches, no card. Hosted MCP at https://mcp.exa.ai/mcp.
  15. 15Exa, Billing and rate limits. Checked 12 November 2026. /search is 10 QPS. Deep types are 5 QPS. Adding $1,000 in credits inside 30 days raises the team to 25 QPS for 90 days.
  16. 16Brave, Search API plans. Checked 12 November 2026. Search is $5 per 1,000 requests at 50 queries a second. Answers are $4 per 1,000 plus $5 per million input and output tokens, at 2 queries a second. A free monthly credit amount is not printed on this plans section.
  17. 17Parallel, API pricing. Checked 12 November 2026. Search turbo and fast $1 per 1,000 requests. Basic and advanced $5. Ten results included. Further results $1 per 1,000. No free allowance and no rate limit on this page.
  18. 18Firecrawl, Pricing. Checked 12 November 2026. Search is 2 credits per 10 results. Free 1,000 credits, listed as 500 searches. Hobby $19 for 5,000 credits, listed as 2,500 searches. Standard $99 for 100,000 credits, listed as 50,000 searches. /search rate limit 10/min on Free, 100 on Hobby, 500 on Standard.
  19. 19Firecrawl, Search. Checked 12 November 2026. Highlights by default. Cost is 2 credits per 10 results, rounded up. Gov results add no search credits. Alexandria tool discovery is free.
  20. 20SerpAPI, Pricing. Checked 12 November 2026. Free 250 searches at 50 an hour. Starter $25 a month for 1,000 searches at 200 an hour. Production $150 for 15,000 at 3,000 an hour. Only successful searches count.
  21. 21Linkup, Pricing. Checked 12 November 2026. Standard, fast, and flash searchResults $0.005 a call. Sourced answer or structured $0.006. Deep searchResults $0.05. Deep sourced answer $0.055. Professional-email accounts start with $20, topped up to $20 a month. No rate limit on this page.
  22. 22Linkup, FAQ. Checked 12 November 2026. Prints $5 per 1,000 queries for flash, fast, and standard searchResults, $6 with a sourced answer, $50 for deep searchResults, and $55 with a sourced answer.
  23. 23You.com, Billing. Checked 12 November 2026. Web Search API $5.00 per 1,000 calls, up to 100 results. Live full-page extraction $1.00 per 1,000 pages. Cached pages on that add-on are free. Changelog’s newest entry on the docs index header’s linked log was 26 October 2026 when that log was opened the same day.
  24. 24You.com, Pricing. Checked 12 November 2026. Web Search API Standard $5 per 1,000 calls. Free MCP profile: 100 queries a day, no API key, no account, no card. Paid QPS is a custom enterprise item.
  25. 25Jina, API pricing. Checked 12 November 2026. s.jina.ai starts at 10,000 tokens a request. No key is blocked. Free and paid keys are 100 requests a minute. Premium is 1,000. The page does not print a dollar price per million tokens.
  26. 26GitHub REST API, repository and commit lists. Read 12 November 2026. GET /repos/{owner}/{repo} for stargazers_count, and the commits list on the default branch. Organization searches returned total_count 0 for openai, anthropics, google, and googleapis queries named in the text. Commit list: https://docs.github.com/en/rest/commits/commits
  27. 27Firecrawl MCP, commit 3.40.0. Committed 12 November 2026, 03:19 UTC. Message: chore(release): 3.40.0. Repository stargazers_count 7,803 at the same read.
  28. 28Tavily MCP, commit on account-picker docs. Committed 5 November 2026, 16:36 UTC. Message: Merge pull request #241, docs for choosing an account for OAuth. stargazers_count 2,486 at the 12 November read.

Cite or embed

Nothing here is estimated. Printed prices are copied from the vendor page. A figure this page multiplies or divides is labeled computed.

Citation

AI Agents Info. "Best web search tool for Claude and OpenAI agents, November 2026: prices and limits of 13 APIs checked." Last updated November 12, 2026. https://www.aiagentsinfo.com/reports/web-search-tools-for-ai-agents-november-2026.

Embed

<blockquote cite="https://www.aiagentsinfo.com/reports/web-search-tools-for-ai-agents-november-2026">
<p>List prices, free tiers, rate limits, result format, and official MCP servers for 13 web search APIs used by Claude and OpenAI agents. Checked November 12, 2026. Nothing here is estimated.</p>
<footer><a href="https://www.aiagentsinfo.com/reports/web-search-tools-for-ai-agents-november-2026">Best web search tool for Claude and OpenAI agents, November 2026: prices and limits of 13 APIs checked</a>, AI Agents Info, last updated November 12, 2026.</footer>
</blockquote>

Related reading

Report

MCP, Nov 2026 to May 2027

The protocol outlook. The MCP table on this page is stars and last commit for search servers only.

November 10, 2026