70k+ reviews
Build your team of AI Agents
Real support, not bots. Define their role, train them on your knowledge, and let them guide and coach while you stay in control.
Trusted by 15,000+ of the world’s top communities
Design agents with clear tasks and skills — just like hiring teammates.
Customize their name, avatar, tone, and voice to match your brand.
Always-on
Give members around-the-clock support, coaching, and guidance.
Agents with real jobs to do
Set responsibilities like you would with a new hire — and know they’ll deliver on day one.
Create and manage up to 10 unique agents per community
Add personality with a custom name, avatar, and bio
Define the agent’s purpose like a job brief with role, style, and limits so they perform exactly as intended
24/7 support and coaching
Give members instant, personalized guidance without waiting for you.
Train agents on spaces, courses, posts, or custom files
Add suggested questions to spark conversations and guide interactions
Keep answers fresh with admin-only and date-based filters
Full control, without chasing
Stay in the loop with full visibility, smart settings, and alerts when members need a real person.
Use Pause AI keywords to hand off sensitive topics to admins
Assign agents to targeted audiences and decide where members can access them
Manage all agent conversations in the shared AI Inbox
Becky Benishek — Online Community Manager, Engineering Unleashed
Powering communities across the globe
15,000+ creators and businesses around the world trust Circle with their communities.
Frequently asked questions
Who gets access to AI Agents?
AI Agents are included for Enterprise and Circle Plus customers at no additional cost in 2025.
Will AI Agents cost extra in the future?
For most customers, no. In 2026, Enterprise and Circle Plus will continue to include AI Agents with generous usage limits. There may be outlier communities with very high-volume, who will need to purchase additional usage. That said, we expect the number of those communities to be small, and we’ll provide ample notice before any pricing or billing changes.
Can I customize my AI Agent?
Yes, you can train your Agent with your community’s knowledge engine, adjust its tone, and monitor all AI interactions in the AI Inbox.
How many Agents can I have active?
You’ll be able to create up to 10 Agents each trained for different tasks.
For example you could have:
- One AI Agent focused on onboarding new members
- Another dedicated to answering course-related questions
- A separate Agent handling billing and technical issues
Can I adjust the AI’s responses if something isn’t quite right?
Absolutely, you can:
- Edit custom snippets with curated knowledge
- Add or update knowledge engine content to refine accuracy
- Monitor interactions in the AI Inbox and pause AI to jump into the conversation if needed
Can AI Agents handle complex or multi-step questions?
AI Agents are great at answering straightforward questions, but for more complex inquiries, they can guide members through multi-step processes or escalate to a human when needed. You can also set up custom snippets to break down multi-step answers effectively.
How does the AI Inbox work if multiple team members are monitoring AI conversations?
The AI Inbox is a shared team workspace where admins can view all AI interactions in real time and manually take over a conversation if needed.
Will my data be used to train models without my permission?
No. Your data is completely secure, and won't be used by Al/LLM platforms (or by Circle) for training models, or for any purpose other than serving your own Al agent.
Al agents use a technique called Retrieval Augmented Generation. In simple terms:
- The user asks your agent a question
- We search your community's knowledge engine for snippets that can help answer the question
- We pass these relevant snippets to the LLM with the system prompt, your guidance on tone and voice, and a memory of the conversation
- These data points are used by the LLM API to generate a response to the user on the fly. When this happens, note that the LLM provider uses parts of these snippets on-to-fly, but no data is stored on their end
- The drafted LLM answer is presented back to your user
Still can’t find the answer? Go to help center
