Bring your own AI keys
Connect your own OpenAI, Google Gemini or Anthropic Claude key instead of the ENA AI Gateway, and learn the fixed daily AI usage limits.
EnaFeedback's AI features run through the ENA AI Gateway by default and need no setup. If your organization already has an AI contract, or you want to decide which provider your data goes to, you can connect your own API key for OpenAI, Google Gemini or Anthropic Claude.
Required role
Provider connections are part of Integrations and are managed only by the Owner role. The feature switches (sentiment, transcription, translation) are turned on and off under AI preferences by the Owner and Administrator roles.
Supported providers
| Provider | Content supported | Note |
|---|---|---|
| ENA AI Gateway | Text, audio, video | Default; no key needed |
| OpenAI | Text, audio | Default model gpt-4.1-mini |
| Google Gemini | Text, audio, video | — |
| Anthropic Claude | Text only | Default model claude-sonnet-4-20250514 |
Your provider choice affects features
Voice transcription needs audio support and video transcription needs video support. Anthropic Claude handles text only, so transcription will not work with it. OpenAI does not support video; for video transcription, use Google Gemini or the ENA AI Gateway.
Connect your key
Get a key from the provider
Create a separate API key for EnaFeedback in your OpenAI, Google or Anthropic account. A dedicated key makes usage easier to track and to revoke if needed.
Open the settings
Under System → System settings, open the AI provider settings.
Add the provider
Choose the provider type and enter your API key. The key is stored encrypted on the server with AES-256-GCM.
Activate and test
Make the provider active. Then send a test feedback item and check under System → Records → AI invocations that the call succeeded with the new provider.
One active provider
- Only one provider is active at a time.
- If the active provider fails, there is no automatic failover to another one. The AI feature returns no result for that call; the feedback itself is still saved.
- If your key expires or is revoked, AI calls show as failed in the log.
Daily usage limits
AI usage is limited by number of calls. The limits are built in, are the same whichever provider you use, and cannot be changed.
| Feature | Daily limit |
|---|---|
| Sentiment analysis | 20,000 calls |
| Translation | 5,000 calls |
| Transcription (voice and video) | 5,000 calls |
| Public translation from visitor screens | 5,000 per IP |
- You cannot set token or money budgets. Token counts appear in the log but are not used as a budget.
- Transcription accepts audio files up to 20 MB and video up to 50 MB.
- Translations and transcripts are cached for 14 days, so repeated text does not use new calls.
Billing is with your provider
When you use your own key, the cost of each call is billed to your account by the provider. Set spending limits in the provider's own console.
Data and privacy
- The AI provider receives visitor comment text for sentiment analysis, voice and video attachments for transcription, and page, survey and module text for translation.
- The AI invocation log under Records keeps provider, model, latency, characters, tokens and status; prompts and content are not stored.
- With your own key, the contract with the provider is directly between your organization and that provider; assess this transfer in your own KVKK records. See KVKK compliance.
Related articles
Can't find what you need?
Our support team can help with anything from setup to reporting.
Translation and languages
Learn about the four visitor languages, automatic AI translation of your content, and how admins read feedback written in another language.
Audit log
Who changed what, and when? Learn to review the audit, notification delivery, AI invocation and translation logs on the Records screen.