Web search and fetch through Parallel, no setup required

WebContextProvider now ships with a Parallel backend, giving agents access to high-quality web search and page fetch through Parallel's hosted research service. The backend exposes both web_search and web_fetch with compressed markdown output, runs keyless by default for fast experimentation, and supports Bearer authentication or OAuth for higher rate limits and production workloads. Default timeouts are tuned for fetching larger pages, so long-running research calls complete reliably out of the box.

Why it matters: Adding web context to an agent traditionally meant wiring up a search API, building a fetcher, and managing rate limits and timeouts. The Parallel backend collapses that work into a single configuration option, so teams can stand up a production-ready research pipeline in minutes and focus on agent behavior instead of infrastructure.

Learn more in our Parallel MCP agent docs.