Premium WebWright accounts can now ground Wren R1 in the live web without wiring a separate search vendor. Two OpenAI-compatible tools, web_search and web_read, run inside our server-side tool loop on POST /v1/chat/completions.

What shipped

  • web_search — search the live web through our private web tools spoke. Send one query and get ranked results, each with a title, URL, and snippet.
  • web_read — fetch a URL and return clean, readable page text, with an optional sandboxed browser render for JavaScript-built content.
  • Premium-only — WebWright runs web_search and web_read for Premium PAYG callers; the execution is what Premium buys. Streaming is supported: the tool rounds run first, then the finished answer arrives as one text/event-stream. Tools you define yourself — your own MCP servers, for instance — are forwarded untouched on every plan and execute in the client.

Full parameter notes, gating rules, and product positioning live on the new Tools page. Pricing Premium now labels this as Integrated tools included with a link to that page.

Why not Firecrawl or Tavily?

Those products are excellent when you want a standalone crawl or RAG search API next to whatever model you already use. WebWright's bet is different: if Wren is your inference path, research should ride the same key, the same meter, and the same completion request. Tool execution stays on our private network; clients never see a separate search endpoint, and you do not manage a third-party search key for the default tools.

Try it

Upgrade to Premium PAYG (or pick Premium at signup), send a non-streaming chat completion with tools enabled, and let the model call web_search / web_read. Details and contrast with external search APIs are on /tools.

— The WebWright Team