Research Agent
Searches the web to inspect known stories, references, products, and brands to write high-fidelity background notes.
Overview
The Research Agent is the first step in the OpenDirector pipeline. It performs web searches to gather factual context about the topic specified in your prompt — including historical references, brand information, product details, and cultural context. The output is a structured research memo that other agents use as the factual foundation for the entire video production. This ensures your content is grounded in accurate, verifiable information rather than AI hallucinations. The agent uses OpenAI's web_search_preview tool to retrieve current information, cross-references multiple sources for accuracy, and flags any conflicting data points. It produces a structured memo with sections for key facts, cultural context, brand guidelines, and source citations — all of which downstream agents consume without copying source text directly. The memo format includes a confidence score for each fact, geographic and demographic context, competitor analysis when relevant, and recommended visual references that the Art Style and Storyboard agents can use to inform their creative decisions.
Input
User prompt / topic
Output
Structured research memo with verified facts, references, and context
Tools
Web search APIs, fact-checking databases