recommend_next
recommend_next analyzes a completed experiment and suggests what messaging angle or dimension to test next — urgency vs curiosity, discount mention, tone, and other strategy tags marginal attaches to subject lines.
When to use it
- Closing the agent loop after get_results
- Deciding the next subject line A/B test in a long-running campaign program
- Compounding learnings across tagged variants instead of one-off sends
Example
{ "experiment_id": "exp_..." }Full loop
generate_variants → launch_test → get_results → recommend_next → repeat. See the documented workflow.