Preferences

Preferences are rules that apply to every future idea generation, not just one batch. They live on the agent and are read every time the Ideas Engine runs. Use them to encode durable constraints: hook styles, audience, off-limits topics, preferred video formats, required CTAs.

Set a preference

Just send a natural-language instruction to the agent that starts with remember: (or similar). The agent recognizes that as a persistent rule and stores it.

curl -X POST "https://agent.gen.pro/v1/agent/run" \
  -H "X-API-Key: $GEN_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "agent_id": "'"$AGENT_ID"'",
    "message": "remember: always use statement hooks, never question hooks. target women 25-34. never mention competitors by name."
  }'

The agent confirms the preference in the run output and stores it on the agent’s config.

Preferences vs per-batch requirements

	Preferences	Per-batch requirements
Scope	Every future run	This run only
How to set	”remember: …"	"focus on …” / “keep under …”
Where stored	Agent config	Just this one message
Override	New “remember:” or “forget:“	Next run clears it

Examples of good preferences

Hook rules — “always use statement hooks; never pose questions in the first 3 seconds”
Audience — “target women 25-34 in the US”
Brand voice — “warm, specific, never condescending; no corporate-speak”
Format prefs — “90% of ideas should be 10-14 seconds; skip anything over 20”
No-go list — “never mention competitors by name; no political content”
Signature style — “always end with a first-person callback to an earlier line”

Removing a preference

Just tell the agent:

forget: that rule about statement hooks. allow questions in hooks again.

Reviewing current preferences

list all my current preferences

The agent returns the active preference set so you can audit what’s applied.