What we collect, and what we do not.

1. Account data

We collect the email you sign up with, a hashed password, API key metadata, wallet balance, transaction records and the usage metadata needed to bill and support you. We do not store full card numbers. Paystack processes card and mobile money payments.

2. Prompts and responses

• Prompts are not used to train Fikra models. Fikra does not train models.
• Prompts and responses are processed to serve the request and to bill it.
• Responses may be cached for up to 300 seconds for non-ZDR requests, scoped to the account. A ZDR request disables the cache.
• Usage records store metadata such as model, token counts, cost, latency and request id. They do not store prompt bodies.
• Upstream providers may retain data under their own policies. The model page names the provider. A universal zero-retention guarantee is not offered.

3. Retention

• Response cache: up to 300 seconds, non-ZDR requests only.
• Batch files: up to 30 days.
• Usage and billing records: kept for billing, tax and dispute handling.
• Support email: kept while needed to resolve the issue and for a reasonable record afterwards.
• Developer request capture: off by default, gated by an environment flag and key allowlist, refused for ZDR requests.

4. Subprocessors

5. Cookies

We use essential cookies and local storage for sessions, theme preference and CSRF protection. We do not sell behavioural data or run advertising pixels.

6. Your choices

You can request access, correction or deletion of account data by emailing [email protected]. Some billing records must be retained for legal and accounting reasons. ZDR requests can be marked per request or per key; see the trust centre.

7. Contact

Fikra API is operated by Lacesse Ventures, Nairobi, Kenya.
Email: [email protected]