Free Web Search APIs for LLM Agents in 2026

An LLM agent that answers questions about today needs a search call that returns clean text, not a page of ads. Four verified offers in the catalog give you that for free: a monthly credit (Exa, Firecrawl), a one-time token balance (Jina) or a free paid plan for students (Tavily). Every number below was checked on the vendor's own pricing or docs page on 2026-10-20. Need to turn a known URL into text instead? Firecrawl and Jina Reader do that too; the free embeddings API guide covers the next step of a RAG pipeline.

Compare the free web search APIs

The searches-per-budget column is our arithmetic from each vendor's published price per call.

Which one to pick

• Most free searches every month: Exa. $10 a month buys about 2,500 Lite searches, and the free plan reaches every endpoint (search, contents, answer) at 10 queries per second.
• Search plus the full page in one call: Firecrawl's /search can scrape each result into Markdown; budget 2 credits per 10 results plus 1 credit per scraped page.
• Ready-to-read results with no extra step: Jina's s.jina.ai returns the top five results with their content already converted to text. The balance is one-time, so it suits a prototype more than a product.
• Students: Tavily gives verified students four months of its $30 Project plan, 4,000 credits a month.

Steps

Exa

1. Sign up at dashboard.exa.ai; the free credit is added with no payment method.
2. Create an API key and save it as EXA_API_KEY.
3. Search: curl -X POST https://api.exa.ai/search -H "x-api-key: $EXA_API_KEY" -H "Content-Type: application/json" -d '{"query": "latest open-weight LLM releases", "type": "auto", "numResults": 5}'
4. Watch the balance on the billing page; it resets to $10 on the 1st of each month.

Firecrawl

1. Sign up at firecrawl.dev (no card) and copy the API key from the dashboard.
2. Search: curl -X POST https://api.firecrawl.dev/v2/search -H "Authorization: Bearer $FIRECRAWL_API_KEY" -H "Content-Type: application/json" -d '{"query": "latest open-weight LLM releases", "limit": 5}'
3. Add scrapeOptions only when you need each page's content: every scraped result costs another credit.

Jina Reader

1. Open jina.ai/reader; a free API key with 10M tokens is generated for you (the search endpoint needs a key).
2. Search: curl "https://s.jina.ai/?q=latest%20open-weight%20LLM%20releases" -H "Authorization: Bearer $JINA_API_KEY"
3. To read one known page instead, prepend https://r.jina.ai/ to its URL.

Tavily (students)

1. Create a free Tavily account.
2. Email support@tavily.com from your official student address, or send a copy of your student ID.
3. Once verified you get a coupon code: on the Billing page, choose Upgrade under the Project plan and enter it as a promotion code.
4. Search: curl -X POST https://api.tavily.com/search -H "Authorization: Bearer $TAVILY_API_KEY" -H "Content-Type: application/json" -d '{"query": "latest open-weight LLM releases", "search_depth": "basic"}'

When to self-host

When the free budgets are too small, run a metasearch engine yourself: SearXNG (AGPL-3.0) queries many search engines at once and can return results as JSON for an agent. There are no credits, but the upstream engines may rate-limit or block a busy instance, and you fetch and clean each page yourself (a Reader-style step).

Limits and the catch

• Credits are spent per call, not per answer. An agent that searches three times per question uses three calls; size the free budget by agent steps, not by users.
• Only Exa and Firecrawl refill for everyone. Jina's 10M tokens are spent once and shared with its Embeddings and Reranker APIs; Tavily's student plan ends after 4 months.
• Jina's model license. The Reader models (ReaderLM-v2, jina-vlm) are CC-BY-NC 4.0 for self-hosting; read Jina's terms before building a commercial product on the free key.
• Rate limits are per account or key. Firecrawl's 10 searches a minute and Jina's 100 a minute are the ceiling for a whole team, not per user.
• Free tiers change. Offers are re-checked monthly; the verified date on each offer page is the day we last confirmed it. The free LLM API comparison lists the models you can pair with these search APIs.