Product · Tools

Integrated web tools, built in.

Premium callers get web_search and web_read on POST /v1/chat/completions. Tools run server-side on our private network; your app never dials the search stack directly.

See Premium pricing Read the launch post

Available tools

`web_search`

Search the live web through our proprietary search system.

Query one string or up to 20 queries in parallel.
Supports dorking operators (site:, filetype:, intitle:, and more).
Optional deep mode extracts full page content from each hit.
Categories for general, news, images, videos, science, and more.

`web_read`

Fetch and clean the body of one or more URLs.

Multi-tier extraction: direct fetch, headless browser, cache, and archives.
Returns readable text suitable for grounding model answers.
Prefer after web_search when you need the full page, not just snippets.

How it works

Premium PAYG — set payg_tier to premium. Standard callers that request tools receive HTTP 403.
OpenAI tools shape — send function tools (or rely on server-injected builtins) on a non-streaming chat completion. Streaming plus tools returns 400; use stream: false or tool_choice: none.
Server-side loop — WebWright runs up to six model-to-tool rounds against our proprietary tool system. Usage is summed across rounds and metered like any other completion.

Compared with Firecrawl and Tavily

Firecrawl

Firecrawl is a crawl-and-scrape API you integrate into your own agent stack. You pay per page, manage keys, and own the tool-calling glue. WebWright ships web_search and web_read as first-class OpenAI tools on Premium: no third-party key, no separate billing line, and tool execution stays on our private network.

Tavily

Tavily optimizes for RAG search as a standalone SaaS. Great if you want a search vendor next to your model provider. On WebWright, search and read are integrated into the same chat completions path as Wren R1: one API key, one usage meter, premium-gated server-side tool loop, and no separate search endpoint for clients to configure.