Trust
Security and data handling
Browser reports describe your pages and your visitors' sessions on them. We keep them in the United States for your plan's retention window, and share them only with the subprocessors below and the integrations you connect.
What we store
Amazon S3
Raw CSP violation reports, as the browser sent them
Each report is stored with your tenant and site ids, the arrival time, the user agent and the IP address it came from, partitioned by tenant, site and date. Apart from rate limiting, which holds it for two minutes and logs it when it refuses a request or fails, this is the only place a visitor's IP address is kept. The other report types skip S3 and go straight to Tinybird. On the Professional plan and above, a second bucket holds verified copies of third-party scripts from sites with payment pages, named by their SHA-256. Only the site's owners and admins can download one.
Tinybird
One analytics row per report, which the dashboard queries
The page, what was reported and where, the source file and line, the policy in force, the referrer and the user agent, and for every type except CSP violations the report body. A network error row records the address of your server the browser was trying to reach. No visitor IP address is written here. Script reports carry a number from 0 to 255 derived from the sender's network, which lets alerts tell one sender from several and cannot be traced back to an address. The free site scanner adds one anonymous row per scan, holding the host, the grade and the finding counts, with no path.
Neon Postgres
Your account, not your reports
Organisations, users and roles, sites and the hosts they accept reports from, plans, alert settings and alert history, and the scripts your sites have reported. On the Professional plan and above, it also holds the daily scans of pages you mark, with cookie values removed, and each decision to authorize or reject a script with its justification and who made it. Decisions cannot be edited while the site exists. A Slack bot token, if you connect Slack, is stored encrypted. Every change HeaderHawk staff make to your account, and each time they view your dashboard, is recorded with the reason given, and that record is kept after the account is deleted. So is the record that an email address has used a free trial.
DynamoDB
Operational state that expires on its own
Quota counters, deduplication markers, alert windows, per-IP rate-limit counters that last two minutes, the policies each site's reports quoted in the last 30 days, and free site scanner results for 30 days. None of it holds report content. A record of each script version a site has reported lasts about 13 months after its last report, so a script you still run is not mistaken for a new one, and it is deleted with the site.
What we fetch
HeaderHawk crawls nothing. On the Professional plan and above, each page you mark for scanning is requested once a day as HeaderHawkMonitor, and a redirect off your site's hosts ends the scan. On a payment page, the scan also fetches up to 25 scripts the HTML names, from any public host, as a desktop Chrome browser so a vendor serves the file your visitors get. It keeps each script's hash.
When you ask it to check your headers, on any plan, HeaderHawk requests one page of your site as HeaderHawkScanner and compares its response headers with your setup, reading the headers only. Once, when you add your first site, it also reads that page and grades it as the free site scanner does. Either way it stops at a redirect off your site's hosts. Of what it fetched, it keeps only the policy and reporting headers the page served, so your next step can tell what you have deployed.
When a site with payment pages reports a script version we have not seen, HeaderHawk fetches it once and keeps a copy only if its SHA-256 matches. That is capped at 100 fetches a site a day, and the copy holds no data about your visitors.
The free site scanner and the hash generator's Fetch and hash send only the URL you enter, which our server fetches once.
Where it runs
Everything we operate runs in AWS us-east-1 (N. Virginia), and so does our Postgres at Neon. The analytics store is Tinybird in AWS us-west-2 (Oregon). Report data is processed in the United States, and a different hosting region can be named in an Enterprise contract.
How long it is kept
A daily job deletes reports older than your plan's window from S3 and the analytics store, along with the scans of your monitored pages. Raw reports and analytics rows also expire on their own after 100 days, above the longest plan window of 90 days, in case that job ever stops.
| Plan | Report retention |
|---|---|
| Free | 15 days |
| Team | 30 days |
| Professional | 90 days |
| Business | 90 days |
| Everything else | Kept for |
|---|---|
| Captured script copies | 12 months from capture |
| Application logs | Up to 90 days |
| Request logs for this site and the docs | 90 days, with no IP address or cookie |
| Internal event archive | 30 days, with no visitor IP address |
| Free site scanner analytics | 12 months |
| Account records | Until 30 days after the account closes |
| Billing records | Seven years |
Subprocessors
These services process data on our behalf, and we sell none of it. Integrations you choose also receive data. Slack gets the alerts you route to it, and Google or GitHub see a sign-in you make through them. The site scanner, the daily scan and a header check send a hostname to hstspreload.org to check the browser HSTS preload list.
Last reviewed against the running infrastructure on 2026-11-13.
Amazon Web Services
Compute, raw report storage, operational counters, the event bus, the CDN and every outbound email.
- Data:
- Raw report bodies, captured script copies, operational counters, application logs, and the address each alert or sign-in email goes to. Request logs for this site and the docs carry no IP address or cookie.
- Region:
- US East (N. Virginia)
Anthropic
Claude, the AI model that suggests a name for a third-party host that a proposed policy's catalog does not know.
- Data:
- A third-party host name and the policy directives it was reported under. Never a URL, a page, a visitor, your organisation or which site loaded it.
- Region:
- United States
Neon
Managed Postgres for accounts, sites, plans, alert settings and alert history.
- Data:
- Your email address, organisation, sites and their allowed hosts, monitored pages and their scans, the script inventory, plan and subscription state, alert settings and history, a Slack bot token if you connect Slack, and the record of staff access to your account.
- Region:
- AWS US East (N. Virginia)
Tinybird
The analytics store behind every dashboard query.
- Data:
- One row per report, with the page, what was reported, source file and line, the script sample where your policy asks for one, the policy in force, referrer and user agent. No visitor IP address. Also one anonymous row per free site scan, with the host and grade but no path.
- Region:
- AWS US West (Oregon)
Descope
Authentication: the login flow, sessions, and team membership.
- Data:
- Your email address, your name if you give one, and login history. No password, because there is none to store.
- Region:
- Vendor-managed
Stripe
Subscription billing for paid plans.
- Data:
- Billing email, payment method and invoice history. Card numbers never reach HeaderHawk.
- Region:
- Vendor-managed
Axiom
Application logs from the Lambda functions.
- Data:
- Request logs and error traces with tenant and site identifiers, timings and outcomes. Report bodies are not logged.
- Region:
- AWS US East (N. Virginia)
Umami
Page analytics on this site, the docs and the dashboard, from a tracker script we serve ourselves.
- Data:
- Page views, and product events carrying only an onboarding step or plan name. No email address, name, identifier or advertising cookie.
- Region:
- Vendor-managed
Authentication
There are no passwords. You sign in with a one-time code sent by email, or with Google or GitHub, so the database holds no password hash to leak. Sign-in and sign-up share one flow, so the login page never reveals whether an address has an account.
Social login runs on OAuth clients we own, and the consent screen names HeaderHawk. Sessions are verified on every API request. On the Business plan and above, an organisation can connect its own SAML identity provider and require it for its verified domains.
How it is protected
- In transit
- HTTPS everywhere, including the ingestion endpoint. This site, the docs and the dashboard are served under an enforced Content-Security-Policy.
- At rest
- The raw report bucket is encrypted and blocks all public access. Postgres and the analytics store are encrypted at rest by their providers. Slack bot tokens are envelope-encrypted under a rotating customer-managed KMS key, and only two functions can decrypt them.
- Tenant isolation
- Every API request resolves to one organisation and every query is filtered by it. Raw reports are stored under a per-tenant prefix, and analytics rows are keyed by tenant first.
- Staff access
- To answer a support question, staff can view your dashboard as one of your members. Each view needs a written reason, ends after 30 minutes and is recorded. The view is read-only, enforced by the API. Changes to your account go through separate operations, each recorded with its reason.
- The ingestion endpoint
- Browsers post reports without credentials, so the endpoint is public. It accepts reports only from hosts your site lists and rate-limits each source. A rejected report gets the same answer as an accepted one, so nobody can probe which hosts you list, and your dashboard counts the rejections.
- Infrastructure
- Everything is defined in code and deployed by a pipeline. The compute is serverless functions, so there are no servers to patch, and only the people who operate HeaderHawk hold administrative access to AWS.
Your data
For the reports you send, you are the controller and HeaderHawk is the processor. We use report data only to run the service for you. It is not sold, shared or used to train anything.
Deleting a site stops it accepting reports straight away, and the reports it sent are deleted from S3 and Tinybird within two hours. To delete your account, have report data erased sooner, or make an access or correction request, email privacy@headerhawk.com.
There is no report export, so an access request is answered by hand from the stores above. The files HeaderHawk produces are the payment-page script inventory and each PCI DSS requirement's evidence from the Compliance page, as CSV on the Professional plan and above.
The privacy policy is the governing document.
Compliance and contracts
HeaderHawk holds no SOC 2 report, ISO 27001 certificate or PCI DSS attestation. We answer security questionnaires and will review your DPA. An Enterprise contract can add an uptime commitment and a named hosting region.
Reporting a vulnerability
Email security@headerhawk.com with what you did, what happened and how to reproduce it. It goes to engineering, and you will hear back within 3 business days.
Test only against your own account, do not run denial-of-service attacks or automated scans against the ingestion endpoint, and give us time to fix an issue before publishing it. We will not take legal action over good-faith research.
The machine-readable version is /.well-known/security.txt, per RFC 9116.
Something not covered here?
Send security questionnaires, architecture questions or your DPA through the contact form and engineering will answer.