Configure follow-up suggestions with FollowupConfig

followups on an Agno Agent or Team takes a bool or a FollowupConfig, and the config carries a count range, a separate model and instructions that Agno appends to the follow-up call's system prompt and nowhere else.

from agno.agent import Agent, FollowupConfig
from agno.models.openai import OpenAIResponses
 
agent = Agent(
    model=OpenAIResponses(id="gpt-5.5"),
    instructions="Help with Python documentation. Decline unrelated requests briefly.",
    followups=FollowupConfig(
        max_followups=3,  # up to 3, possibly none
        model=OpenAIResponses(id="gpt-5.5"),
        instructions="Suggest only Python documentation questions.",
    ),
)

Picture a docs assistant that politely refuses to write a sea poem. Its old follow-ups would often offer the poem again. The default follow-up prompt now asks the model to stay inside what the answer covered, to skip anything the answer declined, to ask for sources when information is missing, and to treat the quoted conversation as data. Every agent and team with follow-ups on gets that prompt. On four refusal scenarios in the pull request's live check, suggestions pointing back at the declined topic fell from 26 of 60 to 2 of 60 on gpt-5.5.

The maximum is enforced now as well. We ran the agent offline with stub models, and the follow-up model always returned five suggestions. Each run also shows the count line Agno sent that model.

Before  followups=True                   "Generate exactly 3 follow-up suggestions."  5 returned
Now     followups=True                   "Generate exactly 3 follow-up suggestions."  3 returned
Now     FollowupConfig(max_followups=3)  "Generate at most 3 follow-up suggestions."  3 returned

The minimum is a request in the prompt, so a model can still send back fewer. A Team takes the same forms, and code that sets followups=True with num_followups or followup_model keeps working. Mixing those top-level arguments with a FollowupConfig raises a ValueError. Follow-ups first shipped in built-in follow-up suggestions for agents and teams.

See the cookbook, and learn more about follow-up suggestions in the documentation.