AnySearch

AnySearch is a search layer built for agents rather than for people. The unit of output is not a page of blue links but a structured result an agent can act on, drawn from trusted sources and filtered before it reaches the model's context window.

What it does differently

Smart routing picks the right source for a query instead of hitting one index for everything, which keeps latency and cost down on queries that do not need a full web crawl.

Structured output means results arrive already shaped — extracted fields rather than raw HTML — so the agent spends its context on reasoning rather than on parsing search results.

Anonymous access is a first-class design choice. Queries are issued without a user-linked identity, which matters when an agent is searching on behalf of someone and the search history would otherwise become a behavioural profile held by a third party.

How you integrate it

Three surfaces are supported natively. A REST API for direct calls, an MCP server so any MCP-capable client can add search as a tool with no code, and a packaged Skill distributed through the team's GitHub releases for agents that use the Agent Skills standard. Which one you reach for depends on whether search is a feature of your product or a capability you want your coding agent to have.

Where it fits

If you are building a research agent, a support bot that must cite sources, or any workflow where the model needs current information it was not trained on, the choice is usually between running your own retrieval stack and renting one. AnySearch is the rented option, tuned for agent consumption and positioned against general-purpose search APIs on freshness, source breadth, and privacy.