Bring Your Own Key

The problem with AI tool pricing

Most AI-powered developer tools follow the same business model: they wrap an API call to Claude, GPT, or another large language model, add a user interface on top, and charge you a per-seat monthly fee that bakes in a substantial markup on the underlying AI compute. Some tools charge $20-40 per developer per month. At scale, this means a 50-person engineering team can easily spend $12,000-24,000 per year on a code review tool. In many cases, the bundled pricing is materially higher than direct model usage costs.

The result is that you end up paying twice. You may already have a Claude Pro subscription, a Codex CLI license, or an enterprise agreement with Anthropic or OpenAI that gives you access to the same frontier models these tools use internally. But the tool vendor does not care – they route your code through their own infrastructure, make their own API calls on their servers, and charge you for compute you could have accessed directly through the subscription you already own.

This model also creates vendor lock-in. Because the tool controls the AI pipeline, switching to a different AI provider means switching to a different tool. You cannot simply point your existing code review workflow at a new model – you have to adopt an entirely new product, migrate your configuration, and retrain your team.

What BYOK means

BYOK – bring your own key – is a different architecture. Instead of wrapping AI API calls behind a proprietary cloud service, VibeRails orchestrates AI tools that you install and configure locally with your own credentials. Here is how it works:

1. You install the AI CLI tool. Claude Code, Codex CLI, or both. These are standalone command-line tools from Anthropic and OpenAI that you authenticate with your own API key or subscription.
2. You install VibeRails. The desktop app detects which AI CLI tools are available on your machine and lets you choose which one to use for each review.
3. VibeRails orchestrates the review. When you run a code review, VibeRails constructs the analysis prompts, manages the session, and structures the output. The actual AI inference happens through the CLI tool you installed, using your credentials, billed to your account.
4. You see exactly what you pay for. The AI compute cost appears in your existing Anthropic or OpenAI billing dashboard, at the same rate you pay for every other use of that subscription. VibeRails charges separately for the software, not for the AI usage.

The separation is clean: VibeRails is the orchestration layer, the user interface, and the project management system. The AI providers are the intelligence layer. You pay each for what they provide, with no middleman markup.

Why it matters

The BYOK model has practical consequences that go beyond pricing, though pricing is where the impact is most immediately felt.

• Cost transparency. Every API call made during a VibeRails review shows up in your existing AI provider billing. You know exactly what code review costs because you see the individual inference charges – not an opaque monthly fee that bundles AI compute with software licensing at an unknown ratio.
• No data routing through third parties. Your code goes directly from your machine to the AI provider you have already evaluated, approved, and signed a data processing agreement with. It does not pass through VibeRails servers. For teams with strict data handling requirements, this eliminates an entire category of vendor risk assessment.
• Provider flexibility. Want to try Claude for one review and Codex for the next? Switch freely. VibeRails supports multiple AI backends and the dual-model verification workflow lets you cross-validate findings across different model architectures. You are not locked into whichever provider the tool vendor negotiated a bulk deal with.
• Model currency. When Anthropic or OpenAI releases a new model version, you get access to it through your existing CLI tool update – you do not have to wait for the code review vendor to integrate and ship it. Your reviews automatically benefit from the latest model improvements.
• Enterprise-friendly. Many enterprises already have negotiated volume pricing, custom terms, or dedicated capacity with AI providers. BYOK lets those agreements extend naturally to code review without requiring a separate procurement process for AI compute through a different vendor.

How VibeRails pricing works with BYOK

Because VibeRails does not resell AI compute, its pricing reflects only the value of the software itself: the orchestration engine, the triage interface, the export pipeline, and the project management layer.

• Free tier – 5 issues per review. Enough to evaluate the workflow and see how VibeRails structures findings before committing.
• Pro – $19/month. Unlimited issues, full export capabilities, all detection categories, priority support.
• Lifetime – $299 one-time. Everything in Pro, forever. No recurring fees, no renewals. Pay once, own it.

These prices cover the VibeRails application. AI usage is billed separately by your AI provider at whatever rate your existing subscription specifies. For most codebases, a full review costs a few dollars in API usage – a fraction of what per-seat AI tools charge for the same analysis.

Who benefits most

The BYOK model is not for everyone – if you have never used an AI coding tool and do not have an existing subscription, the setup has one more step than a fully managed service. But for a growing number of developers and teams, it is the obviously correct architecture:

• Teams already paying for Claude or OpenAI. If your organisation has a Claude Pro, Claude Team, or OpenAI API subscription, you are already paying for the AI compute that code review requires. BYOK means you do not pay for it again through a middleman.
• Cost-conscious startups. When every dollar matters, the difference between a $30/user/month AI tool and a $299 one-time license plus a few dollars per review in API costs is significant – especially as the team grows.
• Enterprises with negotiated AI contracts. Large organisations that have volume pricing agreements with Anthropic or OpenAI can use those rates for code review without a separate procurement cycle. The AI spend stays within existing budget lines and vendor relationships.
• Developers who want pricing predictability. With BYOK, you control the AI spend directly. You can see exactly how much each review costs, set API usage limits in your provider dashboard, and budget accordingly. No surprise invoices from a tool vendor whose usage-based pricing tiers are opaque.
• Security-conscious teams. BYOK means your code is transmitted directly to the AI provider you already trust, using credentials you control. There is no additional vendor in the data flow to evaluate, audit, or worry about.

Get started

The BYOK model extends beyond cloud AI providers. For teams that need complete data isolation, VibeRails also works with fully local AI models. Set ANTHROPIC_BASE_URL to point at a local model server like Ollama or vLLM, and the entire review pipeline runs without any data leaving your machine. See the local AI code review guide for setup instructions.

Install VibeRails, point it at a locally installed AI CLI tool, and run your first review. The free tier gives you 5 issues per review with no credit card required. If you already have Claude Code or Codex CLI installed, setup takes under two minutes.