Full BYOK Support & AI Credit Transparency

May 14, 2026

This release completes a major overhaul of how EnginifyAI handles AI usage for both platform credit users and Bring Your Own Key (BYOK) users. Every AI-powered feature: Builders, Analyzers, Workflows, the Designer, Intent Builder, and the URL-to-Prompt Kit, now enforces the correct usage path based on your account configuration, and gives you clear, accurate feedback after every AI action showing exactly what was consumed and how.

Bug Fixes

• BYOK users were being incorrectly charged platform credits: If you had a BYOK API key configured for a specific provider, several features were still deducting platform credits instead of routing through your key. This affected Builders, the Designer, the Intent Builder, Workflows, and the URL-to-Prompt Kit. The root causes have been identified and corrected across the entire platform, when you have a BYOK key configured for the active provider, your platform credits are no longer touched.
• Certain OpenAI models were failing entirely: Newer OpenAI models that route through the Responses API, including GPT-5 variants, o1, o1-pro, o4, Codex, and computer-use models, were returning errors and producing no output. This was caused by a parameter wiring issue in how EnginifyAI communicated with the OpenAI API. These models now generate correctly.
• Workflow summary generation was missing quota enforcement: The generate summary feature in Workflows had no pre-flight usage check, meaning it would attempt to run regardless of your credit balance. This has been corrected, it now properly checks your quota before running and respects your BYOK configuration like every other AI feature.
• Trial users were not getting BYOK access: Users actively trialing EnginifyAI were not receiving BYOK access at the Explorer level during their trial period, even though paid Explorer subscribers have this access. Trial users now correctly receive Explorer-level BYOK access for the duration of their trial.

Improvements

• Contextual AI usage toasts across every feature: After every AI action, you now see a notification tailored to how the request was actually processed:
  • BYOK users see the number of tokens processed with your API key (e.g., “🔑 4,200 tokens processed with your API key”) Platform credit users see how many credits were used and your remaining balance (e.g., “📊 12 AI credits used • 488 / 500 remaining”)
  These toasts now appear consistently after every generation across Builders, Analyzers, Workflows (step execution, step generation, and summary), the Designer, Intent Builder, and the URL-to-Prompt Kit. Previously many of these surfaces either showed a generic success message or no credit information at all.
• Credit amounts display as whole numbers: Credit usage and balances now show as clean integers throughout the platform rather than decimals, you’ll see “12 AI credits used” instead of “12.00 AI credits used.”
• Consistent notification style across all pages: The Designer was using a legacy notification style while every other page had already moved to the current standard. All AI action feedback now looks and behaves the same across the entire platform.
• Usage limits now track lifetime totals: Quota checks for libraries, prompts, prompt versions, and workflows are now calculated against your cumulative lifetime usage rather than a resetting monthly window. This means usage limits reflect the true total you’ve created since joining, which is how the limits are defined on each plan.
• URL-to-Prompt Kit now uses standard notifications: The URL-to-Prompt Kit previously used browser-native alert dialogs for error messages and had no feedback at all for successful generations. Both have been replaced with the standard notification system used across the rest of the platform.

New Features

• Model selector added to URL-to-Prompt Kit: The configure screen in the URL-to-Prompt Kit now includes a model selector, letting you choose which AI model processes your URL before you run it. Previously the model was fixed and couldn’t be changed from this screen.

Infrastructure

• Reduced database load per AI request: Each AI generation previously required three separate database calls to calculate and record credit usage, one before the generation to estimate cost, and two more after completion to record it. These have been consolidated into a single call by threading the calculated value through the generation pipeline, reducing database overhead on every AI action across the platform.