How are people managing AI coding costs today?

While building Persīv, one thing that kept coming up was how opaque AI coding costs can be.

Many AI coding tools route requests through their own proxy layer. That makes onboarding easy, but it can also mean:

• markup on token usage

• limited visibility into costs

• less control / privacy

With Persīv we tried a different approach — developers connect directly to providers like OpenAI or Anthropic using their own API keys, so you pay the provider price and keep full control over usage.

We also added a small 1.7B local model with tool abilities so basic coding tasks can run completely offline.

I’m curious where others have landed with their workflow:

• Are you mostly using AI coding tools with built-in pricing?

• Bringing your own API keys?

• Or experimenting with local models?

Would love to hear how people here are approaching it.

persiv.ai