I've been evaluating Consensus for our team's workflow over the last two weeks. The core idea is solid—having a single pane to query across multiple models is a game-changer for validation and reducing hallucinations. The execution, however, feels like it's priced for a different reality.
My main issue is the per-user, per-month seat license on top of the model costs. At my scale (a platform team of 8 engineers), the bill quickly becomes astronomical. We're already paying for AWS Bedrock/Titan and OpenAI API usage directly; adding a hefty per-user fee for the orchestration layer makes the whole thing hard to justify. It feels like we're being charged twice.
A more practical model for ops teams would be usage-based, perhaps with a small platform fee. Something like:
- Base platform fee: $X/month for up to Y queries
- Additional queries: $Z per 1k tokens processed (or per query)
- Bring your own keys for model providers
This is how we structure most of our internal tooling costs. The current pricing feels like it's aimed at enterprise sales cycles with unlimited budgets, not at technical teams who need to optimize every dollar.
Has anyone else run the numbers and come to a similar conclusion? Or found a workable configuration that doesn't break the bank? I'd love to be wrong about this.
terraform and chill