Building content with AI agents and the curiosity graph

An AI writing agent with no grounding will happily invent a plausible-sounding article about anything. The problem is that "plausible" and "what your market actually asks" are different sets. CuriosityGraph closes that gap: before the agent writes, it researches the real, ranked questions for the topic and works from those.

REST API. Chart a topic with a job, poll for the result, then read the ranked clusters and tags. Warm topics answer instantly from cache. See the API docs for the full surface — /v1/jobs, /v1/topics/{slug}/cached, /v1/expand, and /v1/briefs.

MCP. If your agent speaks the Model Context Protocol, point it at https://curiositygraph.com/mcp with your API key. It exposes five tools — research_topic, get_tree, get_tags, expand_question, and generate_brief — so the model can pull real questions mid-conversation without any glue code.

The pattern that works: research_topic to get the ranked questions → pick a cluster → expand_question to pull its sub-questions → generate_brief to get a structured outline grounded in real demand → then let the model draft against that brief. Every section maps to a question people actually search, and every priority call is backed by a real volume number rather than the model's imagination.

The live report below is exactly what the research_topic tool returns for one topic — real questions, real volumes.