What data the AI reads, and how Klik Flow protects it.
Klik Flow MCP is designed on the assumption that chat content may contain malicious instructions. The server holds that risk itself, without relying on the AI to stay alert.
Customer text is data
Messages, comments and form entries from customers reach the AI in fields marked untrusted, cut at 1,500 characters, with invisible characters removed. The server instructions tell the AI not to follow instructions in that text.
Personal data
Phone numbers follow the role's masking. Results are capped at 25 items per call, there is no bulk export tool, and results contain no links built from data.
Actions towards customers
Replies go only to chats the business owns (by chat ID, not a free-form number), under the same channel rules as staff. Contacts who unsubscribed or are blocked are never messaged. Messages with links outside the business's addresses, or with phone numbers, are flagged on the draft card and in the confirmation.
Accounts and tokens
Sign-in happens on the Klik Flow page; the AI app never sees the password. Tokens last one hour and work only for the Klik Flow MCP server; tokens in the database are stored hashed and encrypted.
Indonesia's PDP law and AI providers
The business is the data controller and Klik Flow its processor, under Indonesia's Personal Data Protection Law (UU PDP). Data the AI app reads is processed by the AI provider you choose, often outside Indonesia, under its terms.
- Meta's terms for WhatsApp do not allow its platform data to be used to train AI models. Use an AI account that does not train on your data.
- The owner can keep certain chat content out of what AI apps read, in Settings › AI Connection.
- Privacy policy: https://klikflow.id/p/privacy-policy-en (Indonesian original: https://klikflow.id/p/privacy-policy).
- Terms of service: https://klikflow.id/p/terms-en (Indonesian original: https://klikflow.id/p/terms).
Reporting a vulnerability
Found a security vulnerability? Send a report to the contact in https://klikflow.id/.well-known/security.txt or through the Contact page. Do not test it on another business's data.
- Include the steps to reproduce it, when it happened, and the address used.
- Give the Klik Flow team time to fix it before you make it public.