Cloudflare’s Radar Researcher Lets You Ask the Internet Anything in Plain Language

How it works under the hood

Radar Researcher is built entirely on Cloudflare's own developer platform, making it a dogfooding showcase of the company's stack. Here is how each part of the system works:

• Session management: A Cloudflare Worker running the Cloudflare Agents SDK handles each conversation. Every session is a stateful Durable Object with its own SQLite database, preserving chat history and streaming responses even if the user leaves the page .
• AI model: The core reasoning is powered by Workers AI running open models like Kimi K2.7. The model translates a natural language query into a data investigation: looking up relevant Radar datasets, generating interactive charts (line graphs, maps, histograms, etc.), and providing a plain-language summary of the findings .
• Side tasks: Smaller, faster models handle peripheral tasks such as writing conversation titles and suggesting follow-up questions, so they do not slow down the main response .
• Contextual awareness: The tool receives the current date and time, as well as basic connection context (IP location, network), allowing it to tailor location-specific answers .

All information is delivered via Radar's API, ensuring the underlying LLM relies on real, verifiable data rather than generating speculative outputs .

Broader context: Agents Week and WebMCP

Radar Researcher was launched during Cloudflare's Agents Week, a company-wide event spotlighting AI agents built on the Cloudflare developer platform . The entire tool runs on Workers, Durable Objects, Workers AI, AI Gateway, R2 storage, and Cloudflare's data layer — demonstrating the same tools any developer can use to build their own AI agent .

On the same day, Cloudflare also enabled WebMCP support on Radar. WebMCP is an emerging web standard that lets browser-based AI agents discover and call a page's tools directly, instead of scraping HTML. This makes Radar agent-ready — it now passes its own agent-readiness check — and signals Cloudflare's push to make its properties accessible to both humans and autonomous AI agents .

What this means for data access

Radar Researcher lowers the barrier to one of the largest public repositories of Internet measurement data. Previously, extracting insights from Cloudflare Radar required navigating complex visualizations, understanding API endpoints, or manually filtering datasets. Now, a journalist covering a regional outage or a network operator investigating traffic anomalies can simply ask a question and get an answer with supporting charts .

The tool preserves access to the underlying charts, datasets, and query traces, so users can still verify and dig deeper into the findings . Its conclusions remain bounded by what Cloudflare's network and related data sources can observe .