PromptPinchYour customers' questions stay yours.
PromptPinch handles the requests and responses between your AI apps and your AI providers. Here is exactly what we store, for how long, and who can see it.
How we handle your data
Short answers to the questions security reviews ask first.
| Area | What we do |
|---|---|
| Provider API keys | Never stored. Your app sends its key with each request, we use it from memory to forward that request, and it is gone when the request ends. |
| Prompts and answers | Never written to a database or disk. Off by default. If you choose memory savings in your account (confirmed by an email code), an encrypted copy of a response is held in memory for up to 24 hours and wiped whenever our service restarts. You can turn it off at any time. It is never kept for topics you exclude. |
| Request logs | Token counts, model, cost and timing. They never contain message text. A copy is kept for one year in a protected archive in Mumbai, India, so incidents can be investigated. |
| If something goes wrong | We tell affected customers as soon as we are certain, and report to authorities within legal deadlines, including 6 hours for CERT-In in India. We do not hide incidents. |
| Isolation | Each customer's data is kept separate. Nothing stored for your apps is ever used for another customer. |
| Model training | We do not train or fine-tune any model on your data. |
| Encryption in transit | All traffic to PromptPinch and to AI providers uses TLS. |
| Deletion | Nothing about your messages is stored to delete. When you leave, we delete your account data within 30 days; security logs are kept for their one-year period. |
Controls you set
Keep sensitive conversations out of storage entirely.
Excluded topics
Conversations about topics you choose, such as refunds or account details, are never kept, not even in memory.
Per-request opt-out
Your app can mark any single request as private, and it is not kept even in memory.
Memory lifetime you choose
Set how long a response may be held in memory, from one day to never. A restart always clears it.
Subprocessors
The companies that process data on our behalf. We notify customers before adding new ones.
| Company | Purpose | Location |
|---|---|---|
| Railway | Application hosting and databases | See our data processing agreement |
| Cloudflare | Network delivery and protection | Global |
| Amazon Web Services | Protected archive of usage logs (no message text) | Mumbai, India |
| OpenAI or Google | Meaning-matching, only if you switch on similar-question savings | United States |
| Stripe | Billing and payments | United States |
Your chosen AI providers process your requests under your own agreements with them.
Security questions
Do you train models on our data?
No. We never use your prompts or answers to train or fine-tune any model.
Where is our data stored?
Your prompts and answers are not stored anywhere: they are handled in memory. Account and usage numbers are kept with our hosting provider, and a one-year archive of usage logs (no message text) is kept in Mumbai, India.
Can we exclude sensitive topics from storage?
Yes. You choose topics that are never kept, even in memory, and your apps can mark any single request as private.
Do you offer a data processing agreement?
Yes. It is available on request for every plan and included with Enterprise.
Need our security documents?
Request our data processing agreement and security questionnaire answers.