Frequently Asked Questions
Common questions about Crucible's features, pricing, and capabilities.
How accurate are the debate outcomes?
Crucible forces Agents to challenge each other's claims, burn away errors, and converge on defensible conclusions. You receive a brief with full citations, areas of dissent, and confidence scores. Every brief is built for scrutiny—we still recommend a human strategist reviews before execution.
Can I control how much I spend per session?
Yes. You deploy the number of Agents you need and select the intelligence tier for each seat. The pricing calculator on our pricing page shows the total cost before you start. Enterprise workspaces can set per-session caps or require approvals.
What happens to my data after a session?
Session events and debate content are stored securely to enable history and analysis. You configure retention periods in settings, and the system automatically purges data according to your preferences. You can delete individual sessions or your entire account at any time. See our privacy policy for details.
Which models and providers do you support?
Crucible orchestrates premium intelligence from OpenAI (o3), Anthropic (Claude Sonnet 4.5), DeepSeek (R1), Google (Gemini 2.5), MiniMax (M1), and xAI (Grok 4). Assign models per Agent or let the system auto-seat the optimal mix for your challenge.
Are there limits on session length or team size?
Sessions typically complete within 45 to 60 minutes, including convergence. Each simulation can host up to twelve Agents, and you can run multiple sessions in parallel.
Can I use my own AI models?
Yes, with the Community Edition. You can self-host Crucible and connect your own API credentials for any supported model provider. Your prompts and data never leave your infrastructure.
How do I export session artifacts?
Decision Briefs and Meeting Minutes can be exported as PDFs directly from the session view. API access for programmatic export is coming soon. Enterprise customers can configure automatic exports to their workspace.
What's the difference between Agents and Knights?
"Agents" is the current terminology. "Knights" was the previous term—you may still see it in some older documentation or code. They refer to the same concept: specialized AI expert personas that participate in debates.
What is your quality guarantee and refund policy?
We stand behind the quality of every session. If you're not satisfied with your debate outcome, you can request a full refund within 24 hours of session completion, no questions asked. Simply click the refund button on your completed session to initiate the process. We track refund patterns to prevent abuse and reserve the right to ban accounts that repeatedly abuse this policy. See our Terms of Service for complete details.
How do I request a refund if I'm not satisfied?
If you're not satisfied with your session, you can request a refund directly from your completed session page. Look for the refund button in the session header, it's available for 24 hours after completion. The refund process is simple: select a reason, submit your request, and if approved, your refund will be processed automatically. Refunds are limited to one per user per 30 days to prevent abuse. For technical issues or payment problems, contact our support team.
Still have questions? Email us at hello@roundtablelabs.ai or check out our documentation.