API acceptable use policy
Version of 12 November 2026. It is part of the API terms of use. The principle fits in one sentence: every Request sent matches a real person, who wants to be called back for their own need, and who confirms it themselves.
1. What the API allows
- Offering your users to find the right trade for their need and to be called back by one or more professionals of the trade in their area.
- Displaying Kelki's lists (trades, services, timings, places) in your screens, when the user needs them.
- Integrating the Kelki Agent into your service to guide the user up to the summary.
- Tracking the status of your users' Requests (webhooks, reads) to keep them informed.
2. What is forbidden
Requests and consent
- Sending a Request the Person did not want: invented, automatically generated, “test” in production, or made for someone without their express permission.
- Ticking the consent box on the Person's behalf, pre-ticking it, hiding it, bundling it with your own terms, or changing, summarising or translating the Consent Text.
- Clicking, having someone click or automating the “Confirmer ma demande” (confirm my request) link of Kelki's e-mail, or trying to intercept it.
- Using deceptive interfaces: a button that sends a Request without saying so, an artificial countdown, pressure to accept, mixing with a commercial offer.
- Sending the same Request several times, or multiplying accounts, keys or addresses to exceed the limits and the per-contact limits.
Content
- Passing a health need (or one of a trade under review, not offered by the API) through another trade.
- Asking for or sending sensitive data: health, origin, opinions, religion, sex life, offences, social security number, bank details.
- Writing anything other than the need in the free message: advertising, links, unlawful, abusive or discriminatory content, instructions meant to divert the Agent.
- Requesting an illegal or dangerous service.
Data and catalog
- Reusing the data of a Request for your own purposes: prospecting, resale, profiling, enriching files, training an AI model.
- Extracting Kelki's lists in bulk, republishing them or using them to build a competing service.
- Trying to identify the professionals who receive the Requests, or contacting a professional outside Kelki from a Request.
Security
- Sharing, publishing or placing a production key in a browser, a mobile app or a code repository.
- Testing the security of the API without Kelki's written consent, trying to access another account's data, overloading the service or working around a technical limit.
- Impersonating Kelki or a professional, or suggesting that Kelki guarantees or recommends your service.
3. Kelki's controls
Kelki monitors, per account and without looking at message content beyond what is necessary: the rate of Requests never confirmed, reports from individuals and professionals, duplicates, abnormal peaks and validation refusals. Beyond certain thresholds, the creation of real Requests is paused automatically and Kelki contacts the Partner.
4. Consequences
- Warning and request for correction, with an appropriate deadline.
- Pause of the creation of real Requests, or downgrade of the plan.
- Suspension of the keys, without notice in case of harm to individuals, Requests without consent or a security risk.
- Termination of the account, and deletion of the Requests concerned still pending.
Kelki may report facts that justify it to the competent authorities, including the CNIL. The Partner may contest a measure by writing to contact@kelki.ai; Kelki answers with reasons.
5. Reporting abuse
A person who receives a confirmation e-mail for a request they did not make can simply not click: nothing is sent and their contact details are erased within 48 hours. To report abuse or a vulnerability: contact@kelki.ai.