Appropriate uses and limits for Pollen's privacy-safe aggregate prompt intelligence.

Use Cases

Pollen is useful when aggregate behavior is more informative than an individual anecdote.

MCP ecosystem research

Track which public MCP servers and tools appear in privacy-qualified cohorts, how their published volume changes, and which outcome buckets follow. Use this as one market signal rather than as a count of all installations.

Agent and model evaluation

Compare aggregate model usage, tool-category sequences, terminal states, and check results. Pollen shows observed correlation, not causation, and does not reveal an individual contributor's workflow.

Product research

Use coarse intent and workflow trends to identify areas worth interviewing users about. Pollen is not a substitute for qualitative research, and a missing cell can mean the cohort is below k=5 rather than demand is zero.

Research limits

• No raw prompt or session access
• No proof that a human authored each prompt
• No individual-level targeting or attribution
• No below-threshold cohorts
• No guarantee that the founding panel represents the broader developer population

Check /catalog and /network before purchasing history so you can evaluate coverage, freshness, and methodology.