Bringing Your Own AI Subscription: No API Costs

Most AI platforms charge per token. Every prompt, every response, every retry adds to your bill. For workflows that run frequently or process large amounts of text, API costs can grow quickly and unpredictably. FlowKoi takes a different approach: you bring your own Claude subscription.

How the BYOS Model Works

FlowKoi runs Claude CLI inside Docker containers. Claude CLI authenticates using your existing Anthropic subscription — the same one you use for Claude.ai or Claude Desktop. Your subscription credentials are passed into the container via OAuth tokens, and the CLI uses your account’s quota directly.

This means FlowKoi does not sit between you and Anthropic as an API proxy. There is no markup on tokens, no per-request billing, and no surprise invoices at the end of the month. You pay Anthropic for your subscription and you pay FlowKoi for the platform. The two costs are completely separate.

Why This Matters Financially

Consider a workflow that runs daily and processes roughly 50,000 tokens per run (a moderate amount for a code review or report generation task). At typical API rates, that is approximately $1.50 to $3.00 per run, depending on the model and the input/output ratio. Over a month, that single workflow costs $45 to $90.

Now multiply that by ten workflows. You are looking at $450 to $900 per month in API costs alone — before you pay for the platform, compute, or storage.

With a Claude Pro or Team subscription, you get a generous daily usage allowance for a fixed monthly price. Running the same ten workflows through Claude CLI on your subscription keeps your AI costs flat and predictable.

The Team Subscription Advantage

For organizations, Claude’s Team and Enterprise plans offer pooled usage across seats. A team of five can share a subscription and run workflows under a shared account. This is particularly useful when workflow execution is concentrated — most teams have a few heavy workflows and many lightweight ones.

FlowKoi supports configuring different credentials per workflow, so you can assign specific workflows to specific subscription accounts if you need to track usage separately.

When API Billing Still Makes Sense

The BYOS model is not right for every situation. If you are building a product that makes Claude API calls on behalf of your end users, you need API billing — you cannot share your subscription credentials with customers. Similarly, if you need fine-grained usage tracking for chargeback or cost allocation, the API’s per-request billing gives you that granularity.

But for internal workflows, dev tooling, and team automation — the scenarios FlowKoi is designed for — using your existing subscription is almost always the more economical choice.

Setting Up Your Credentials

Getting started is straightforward:

1. Sign into Claude CLI on your local machine using claude login.
2. Copy the resulting credentials (found in your Claude config directory) into FlowKoi’s credential configuration for your workflow.
3. FlowKoi injects these credentials into the container at runtime, and the CLI authenticates automatically.

Your credentials are stored encrypted and scoped to your account. They are injected into containers at runtime and destroyed when the container shuts down. No credentials are persisted in container images or logs.

Predictable Costs, Better Planning

Flat-rate AI costs change how teams think about automation. When every API call has a price, there is a natural hesitation to experiment. Teams optimize for fewer calls instead of better results. With a subscription model, you can iterate freely — run a workflow five times to get it right without worrying about the bill.

The best automation is the kind you do not have to think about. Predictable costs make that possible.