Server tools
Give models access to web search, code execution, and more. Explore what each tool does, compare providers and pricing, and find out how to use it. Read the server tools guide. Workspace admins can manage available server tools from the Tools page in workspace settings.
All tools
Frequently asked questions
Server tools are built-in capabilities that OpenRouter exposes to any tool-calling model during a request, including web search, web fetch, image generation, a sandboxed shell, and orchestration tools. For hosted server tools and engines, OpenRouter executes the tool loop server-side so your application does not need to host tool backends; client-executed tools such as default bash and apply_patch return structured tool calls for your application to run locally.
Most server tools work with any tool-calling model through Chat Completions (/api/v1/chat/completions), the Responses API (/api/v1/responses), or the Anthropic Messages API (/api/v1/messages). A few tools follow one API shape, such as apply_patch on Responses and bash on Anthropic Messages, as noted on each tool page.
In-process utilities like datetime and tool search have no separate tool fee beyond normal model token usage. Orchestration and media tools such as advisor, subagent, fusion, and image generation are billed at the underlying models’ token or image prices, while hosted search, fetch, and sandbox tools list per-engine pricing on each tool page.
Yes. Workspace admins can enable, restrict, or lock allowed engines for server tools from the Tools page in workspace settings, and per-request data policies such as Zero Data Retention (ZDR) restrict engines that do not meet that policy.