Security & data
What happens to your prompts, completions, API keys, and metadata, including where each is processed and how long it's kept.
Prompts and completions pass through NeuroAPI, are billed, and are not retained. The only data NeuroAPI keeps is what it needs to authenticate, meter, and debug, and even that excludes the content of your prompts. The next section lists the two exceptions.
What NeuroAPI doesn't store
- Prompt text. Not persisted in the NeuroAPI database, not written to logs, and not kept after the response is delivered. When a data source fails, the search query the agent built from your prompt can appear in NeuroAPI's operational logs.
message_historycontent. Same as prompts.- Completion text. Streamed or returned, then discarded. One
exception: when you send an
Idempotency-Key, the response is kept for 24 hours so a retry with the same key gets the same answer, then deleted. - Training corpora. Neurobro does not use your traffic to train any model. Besides the model provider serving the request, the agent may send search or data queries derived from your prompt to the data sources it uses to answer (for example web search and market data).
What NeuroAPI stores
| Scope | Fields |
|---|---|
| Per call | request_id, timestamp, API-key prefix (neuro_<8 chars>), HTTP status, cost_units, mode, request and response byte counts |
| Per account | account email, name (if given), hashed password, Stripe customer / subscription ids, monthly usage counters, list of API-key prefixes |
The API-key prefix is the public identifier (e.g.
neuro_a1b2c3d4) - never the full secret. The full secret is shown
once at creation and never again; see API-key
handling for how it's stored.
Where prompts go
Prompts and completions are processed by a mix of two backends:
- Third-party model providers - Anthropic, OpenAI, and DeepSeek - under their respective enterprise data-handling terms, which exclude API traffic from training corpora at the time of writing.
- Open-source models and Neurobro's own fine-tuned models, hosted on Neurobro infrastructure. Traffic served by these models stays entirely within Neurobro and is not shared with any third party.
The mix is selected per request based on the chosen mode. To exclude
a specific third-party provider, or to keep your traffic on
Neurobro-hosted models only, email [email protected] and NeuroAPI
support will set up your account to match.
Logging policy
- Application logs record the API-key prefix only; full secrets are never written. Access logs mask credentials and search terms in URLs.
- Error reports sent to NeuroAPI's error-tracking provider never include request bodies, query strings, cookies, or local variables, and known secret formats are redacted from any text they contain.
- The
X-Request-Idheader on every response is the safe correlation id to share with support - it does not encode any account or key material. - Operational logs are kept for the minimum window needed to debug and meter usage.
API-key handling
- At rest. Keys are stored as a peppered HMAC-SHA256 hash, which can't be turned back into the key. Plaintext exists in memory only at creation time and is discarded after it's shown to you.
- On the wire. TLS 1.2+ is enforced at the edge.
- Rotation. Mint a new key, deploy alongside, switch traffic, then
revoke the old one - see the Authentication
guide. Revoked keys return
401on the next request; there is no grace period. To stop every key at once, see Revoke every key at once. - Misuse protection. Repeated failed-auth attempts from the same source are rate-limited (20 / minute) to throttle credential guessing.
Encryption
These mechanisms protect your data:
| Layer | Mechanism |
|---|---|
| In transit | TLS 1.2+ on every external connection |
| Database backups | Encrypted at rest |
| API keys | Peppered HMAC-SHA256 hash |
Data retention
| Data | Retention |
|---|---|
Prompts, completions, message_history | Not retained |
Responses cached for an Idempotency-Key | 24 hours |
| Operational logs (request_id, status, prefix, byte counts) | Minimum needed to debug + meter |
| Billing / usage records | Retained for tax and accounting requirements |
| Account record | Until you delete the account, then purged on next batch |
Compliance & data-subject requests
- GDPR. Data-subject requests (access, deletion, export) go to
[email protected]. - DPA. A Data Processing Addendum is available on request to the same address.
Reporting a vulnerability
Email [email protected]. Before you test, read the Neurobro
Vulnerability Disclosure
Policy. It
covers NeuroAPI (neuroapi.neurobro.ai and api.neurobro.ai) and
sets out what is in scope, the testing rules, safe harbor, and how
fast Neurobro responds.
Denial-of-service and load testing are not authorized. There is no paid bounty program. Neurobro credits reporters when a fix ships, with their permission.