How can I combine multiple AI model subscriptions and run a cross-model “voting” system for best-answer generation?

[igelmigel4]

I want to build a system where several AI models receive the same user query, generate their own answers independently, and then collectively “vote” or compare their outputs to determine the best final answer.

However, I only have standard consumer subscriptions for these products (not API plans):

• ChatGPT Plus
• Perplexity Pro
• Claude Pro
• Grok Pro
• Gemini Pro

My challenge:
These consumer subscriptions normally do not provide direct API access, which makes automated querying or orchestrating multiple models difficult.

My questions
Is it technically possible to implement a multi-model pipeline using these non-API subscriptions?
Are there any legal/ToS-compliant ways to programmatically send prompts to these models without paid API access?
If not, what would be the minimal architecture or toolset required to build a “model congress” where each model returns an answer, and a separate agent evaluates or votes on them?
Are there recommended open-source frameworks for managing multi-model orchestration and answer evaluation?
I’m looking for guidance on how such a system can be built — or if it's even feasible — given the limitations of standard, non-API consumer products.