Search results without a chat's retrieval layer
FreeSerp, at freeserp.ai, is a free web search and SERP API for AI agents: no key, no signup, no rate plan. It draws on two sources, its own 3.1-billion-plus-page index and a 20-million-plus site-profile database with LLM-written summaries, positioned as a keyless alternative to the Google and Bing search APIs.
When an agent asks a chat product to look something up, the result has already passed through that product's own retrieval choices. A direct SERP API removes that extra layer, for better or worse.
Related tools
On this page
Two sources, not one
FreeSerp Global is its own index of more than 3.1 billion web pages; FreeSerp Main is a database of more than 20 million site profiles with summaries written by an LLM. An agent can query either, depending on whether it needs broad page-level search or a faster profile-level answer about a specific site.
What keyless removes
No API key, no signup and no rate plan sit between a request and a result. FreeSerp positions itself specifically as a keyless alternative to the Google and Bing search APIs, the layer of friction it is built to remove, not a claim about the search results themselves being edited or unedited.
JSON REST API and MCP connector
Requests return structured JSON, and an MCP connector is available for agent frameworks built to call tools directly. Either way, an agent's code does not need bespoke parsing logic for scraped HTML, a separate, mechanical kind of simplification from anything about content moderation, and it means fewer places where a page layout change quietly breaks a working integration.
What is not documented
FreeSerp does not claim to scrape Google directly, and no guarantee of real-time Google-equivalent results, data freshness, or rate limits is published, since none of those are stated features. Treat unfiltered here as a claim about authentication, not as a claim about the freshness or completeness of what comes back.
Where this fits a no-filter agent
An agent that fetches its own search results, rather than relying on a chat product's built-in browsing choices, has one fewer undocumented layer between a question and an answer. That is a narrow, specific gain, not a claim that the web itself becomes unfiltered by using an API instead of a browser.
Choosing between the two sources
FreeSerp Global suits a broad question where the answer could come from any of 3.1 billion-plus pages and you want the agent doing its own reading. FreeSerp Main suits a narrower question about a specific site, since its 20 million-plus profiles already carry an LLM-written summary, trading some depth for speed. Neither source is described as filtered for quality beyond what indexing naturally selects, so an agent using either should still treat a single result as a lead to verify, not a settled fact.