Every AI is sometimes confidently wrong. Conclave puts AI models from different labs on one council. They answer your question independently, cross-examine each other, and Claude Opus presides and delivers one verdict: what you can rely on, every error found and corrected, how strongly the models agree, and who dissented. Every verdict gets a signed receipt anchored on public blockchains.
Claude is included: Claude Opus 5.5 presides, with Claude Sonnet 5.5 and Claude Haiku 4.5 as members. OpenAI's GPT, Google's Gemini, xAI's Grok, DeepSeek and Mistral join with an API key, and one OpenRouter key adds open-weight models such as Qwen, Llama, GPT-OSS, Gemma, Kimi, GLM and Nemotron.
On 7 business questions with verified answers (loan math, a pricing trap, S-corporation rules, 2026 retirement limits, a contract clause, a collections email and break-even math), a council of Claude Sonnet 5.5, Grok 4.7, DeepSeek V4 Pro and Mistral Large 4, with Claude Opus 5.5 presiding, was tested on 11 October 2026. Single models gave a wrong number or conclusion in 3 of 26 answers. The Conclave verdict had none: 0 of 7, and it caught all 3 mistakes. On the 2026 retirement-limits question only 1 of 4 models was right, and the verdict sided with it. An earlier run on 10 October (Claude and Mistral council) found 4 of 20 single-model answers wrong and 0 of 7 verdicts. Grok and DeepSeek ran on connected API keys. Each answer was graded by two AI judges; a mistake counts only when both flagged it.
Pro $49 a month for one person. Team $79 per user a month (3-user minimum). Bring your own key $29 per user a month, unlimited on your own API keys. Enterprise from $1,500 a month. There is no free plan; you can try the council check with free calls inside RILBOT.
No AI is. Models can agree and still be wrong. Conclave makes errors much easier to catch: independent answers, cross-examination, a consensus score, and live fact-checks in Verify mode. For legal, tax, medical or large financial decisions, confirm with a qualified professional. Conclave tells you where to focus that check.
Claude is included: Claude Opus 5.5 presides, with Claude Sonnet 5.5 and Claude Haiku 4.5 as members. OpenAI's GPT, Google's Gemini, xAI's Grok, DeepSeek and Mistral join as soon as you connect your own API key for them, and one OpenRouter key adds every lab including the open-weight models (Qwen, Llama, GPT-OSS, Gemma, Kimi, GLM and Nemotron). Conclave automatically seats each lab's newest flagship, and you choose who sits on your council.
A single assistant gives you one model's answer in a confident voice. Conclave gets several independent answers, has them check each other, and tells you how strongly they agree, which errors were fixed and who dissented, with a signed receipt.
Your conversations are saved in your workspace so you can come back to them, and you can delete any of them. To be answered, a question is sent to the AI providers on the council. Conclave does not train models on your data. Receipts contain only fingerprints, never your text.
Pro is $49 a month for one person (150 council answers and 1,000 Swift answers). Team is $79 per user a month with a 3-user minimum (250 council answers per user, pooled). Bring your own key is $29 per user a month, unlimited on your own API keys. Enterprise starts at $1,500 a month. Extra council answers come in packs of 100 for $25. There is no free plan.
Yes, with API keys. A workspace owner can connect the company's own Anthropic key (for Claude) and OpenRouter key (for GPT, Gemini, Grok, DeepSeek and the other models). Those providers then bill you directly and your Conclave budget isn't used. Keys are checked, stored encrypted and never shown again. Consumer subscriptions such as ChatGPT Plus or Claude Pro can't be connected, because their makers only allow them inside their own apps.
Rip It Labs LLC, an Arizona company that builds AI agents, blockchain proof systems and RIL Warden, the gateway Conclave runs behind.
Put the same question to several AI models from different labs, have them check each other's answers, and look for where they disagree. Conclave does this automatically: the models answer independently, cross-examine each other, and Claude Opus rules on a verdict that lists the errors found, the corrected figures and any dissent. You can also paste an existing ChatGPT answer into Conclave's check inside RILBOT (rilbot.ripitlabs.com/conclave) to have the council review it.
An AI council is a group of AI models from different companies that answer the same question independently and then review each other's work, so one model's mistake can be caught by the others. Models trained by different teams on different data tend to make different mistakes, which is what makes a council more reliable than any single model.
Open Conclave · RILBOT AI workforce · RIL Warden · Made by Rip It Labs