A composable research and targeting toolkit: generate one hint, inspect its evidence, map a niche, or investigate an advertiser without leaving the agent conversation.
You do not need to memorize tool names. Give Codex or Claude a campaign question and enough product context; the MCP server instructions explain how the capabilities fit together. Use the directory below when you want to understand or audit the tool chain.
Contextgenerate_context_hintTurn a product brief into a primary hint, alternate direction, targeting breakdown, matched prompts, similar advertisers, and competitor evidence.
You have enough product, audience, and differentiation context to build or pressure-test a campaign direction.
Example requestUse the product brief in this workspace to generate and validate a context hint.
list_nichesReturn the current niche catalog and show which niches have enough captured evidence for deeper intelligence.
You need a valid niche name or want to see where evidence coverage exists before generating.
Example requestWhich available niche is the best starting point for this product?
explain_context_hintsGive the agent the canonical definition, writing guidance, keyword comparison, examples, or strategy principles.
The user needs the targeting model explained before making or reviewing a hint.
Example requestExplain how a context hint differs from a keyword list.
get_niche_intelInspect top advertisers, intent mix, and confidence-gated inferred context hints in one market.
You are researching a market, selecting an intent lane, or checking how crowded a targeting pattern is.
Example requestShow the intent mix and representative hints in SaaS and productivity.
get_advertiser_intelAnalyze one advertiser's observed niche presence and inferred audience, intent, topic, and context-hint patterns.
You need a focused competitor teardown across the markets where that advertiser appears.
Example requestHow does this advertiser appear to target buyers across niches?
find_similar_advertisersFind advertisers whose captured ChatGPT ads are semantically close to a product or campaign brief.
You need a competitive set before deeper niche or advertiser research.
Example requestFind the closest advertisers to this product and explain why they match.
Resolve the market. List or select a niche with enough evidence.
Map the field. Find similar advertisers and inspect the niche or closest competitors.
Generate and validate. Create the primary hint, alternate direction, prompt matches, and quality breakdown.
Interrogate the evidence. Compare prompts and inferred competitor patterns with the actual buyer before accepting the direction.