Trust centre
What we can prove, and what we cannot yet
Everything on this page is checkable. Where a control is designed but not built, or a certification is not held, it says so in the same size type as the good news.
Certifications and attestations
The honest version. A status of “in preparation” means work has started and nothing has been issued; it does not mean nearly done.
-
SOC 2 Type I
In preparationWe do not have a SOC 2 report. Preparation is under way: the control matrix, evidence collection and the security review programme exist, and an auditor has not yet been engaged. There is no report to request, and we will not imply otherwise. This page is updated when an auditor is engaged and again when a report is issued.
-
SOC 2 Type II
Not heldFollows Type I after a six-month observation window. Not started.
-
HIPAA
Not a certificationHIPAA has no certifying body, so nobody can be “HIPAA certified”. We act as a Business Associate under a signed BAA, and only for instances placed as secure hosting. HIPAA hosting opens after the independent penetration test. Until then you can join the waitlist: sign the BAA in the portal or contact sales.
-
Independent penetration test
Not heldNo external penetration test has been completed, and one is not yet commissioned. When it is done we publish the tester, scope, dates, finding counts by severity and remediation status on our security page. HIPAA secure hosting does not open to customers before it is finished and its critical and high findings are fixed.
-
ISO 27001
Not heldNot held and not in progress. We will start it when a customer requires it.
-
PCI DSS
Not a certificationCard details never reach us. Checkout and card storage are handled by Stripe; we hold a billing contact and a payment-method reference.
-
GDPR
Not a certificationA regulation, not a certification. Our data processing agreement is published and incorporates the EU Standard Contractual Clauses, the UK International Data Transfer Addendum and the Swiss amendments.
-
WCAG 2.2 AA
In preparationOur accessibility statement declares the site and portals partially conformant and lists the items that do not yet meet the standard. A VPAT 2.5 report is available on request. No third-party accessibility audit has been completed.
Request our documentation
There is no audit report to send. This is how you ask for what does exist, and how you get told when a report is issued.
What you can have today
Everything on this page, our security architecture page, the data processing agreement, the Business Associate Agreement, the sub-processor list and our accessibility statement are public and need no paperwork.
Under a mutual NDA we will also share an architecture and threat-model summary, the full HIPAA control narrative and the detail behind the sub-processor list. We will not share a SOC 2 report or a penetration-test summary, because neither exists.
How to ask
If you are already a customer, open a support ticket in the portal under the Security category. It is authenticated, it is logged, and any document comes back as a time-limited link inside the ticket.
Otherwise, email [email protected]. Tell us the legal entity, your name and role, a work email on that entity's own domain, which documents you want and what the review is for. We sign our mutual NDA or yours.
There is no form on this page on purpose. A form would mean sending your details to a third party we have not vetted, and a document request that is approved automatically is a document request that eventually goes to the wrong address. A person reads every request and a person approves it.
What to expect
- Answer to your request
- Within 2 business days
- Delivery after a fully signed NDA
- Within 5 business days
- How documents arrive
- A watermarked copy on a link that expires after 14 days, sent only to the address on the NDA
- What we record
- Who asked, for what, which NDA, who approved and when access expires
We may decline a request. You will be told, and told why.
Sub-processors
Third parties that process data on our behalf. The authoritative list, with each provider named and located, is on the docs site; we give at least 30 days' notice before adding or replacing one.
| Purpose | What they receive |
|---|---|
| Cloud infrastructure, object storage for backups, key management | Customer instance data (encrypted) and backups, in the region you choose |
| Payments, invoicing and tax | Billing contact and a payment-method reference; card details never reach us |
| Customer identity | Email address, password hash, MFA factors |
| Transactional email | Email address and notification content; never PHI |
| Edge protection for the public API and website | Request metadata and IP addresses |
| Compliance evidence collection | Staff and configuration metadata; no customer data |
See the full list, with each provider named and its location. Sub-processors with access to infrastructure that can hold protected health information have a Business Associate Agreement with us. To object to an addition, write to [email protected].
Data residency
You pick the region. We do not move your data out of it.
| Region | Status |
|---|---|
| US East (N. Virginia) us-east | Available |
| EU West (Ireland) eu-west | Planned, not yet serving |
- You choose a region per instance. Data stays in that region and is not moved between regions by us.
- Cross-region backup copies exist only where the plan's backup policy enables them, and that is your choice of plan, not a background transfer.
- Organisations with a signed BAA pin a region; backups and copies stay inside the country.
- The control plane orchestrates instances and never reads what is inside them.
- Cross-region backup copies are part of the hourly, long-retention backup policies only.
Encryption
Including the hop that is not encrypted yet.
- In transit, public
- TLS 1.2 or 1.3 is required on every public connection, to the API, the portals and your database endpoints. Optional client certificates (mTLS) per instance.
- In transit, inside a cell
- Our services speak mutual TLS to each other. On the gateway-to-database hop, MySQL and MariaDB use verified TLS; the other engine adapters are currently plaintext on that internal hop, restricted by a default-deny network policy but not encrypted. We would rather tell you than let you assume.
- At rest
- Volumes and backups are encrypted with keys managed per tier, so a free-tier key cannot open paid or HIPAA data. Per-organisation customer-managed keys are part of secure hosting and are not built yet.
- Credentials
- Your database credentials are generated inside the cell and the control plane stores only a reference, never the password. Revealing one is a single-use, signed, short-lived fetch straight from the cell. API keys are hashed with argon2id and shown once.
- Audit records
- Every control-plane and staff action is written to an audit log with a per-day hash chain, exported daily to object storage under a compliance lock and retained for seven years.
Incident response
What happens between something going wrong and you hearing about it.
-
Step 1
Detect
Security-relevant actions across the platform are emitted as structured security events and correlated by our detection service; alerts route to an on-call rotation with escalation and dead-letter handling.
-
Step 2
Classify
Four severities, from SEV1 (customer data at risk, or a region down) to SEV4 (cosmetic). Each alert class has a written runbook; an alert without a runbook is treated as a bug.
-
Step 3
Contain and communicate
Status page updates during the incident, and direct notice to the security contact registered on affected organisations.
-
Step 4
Notify
For personal data, notice to you without undue delay under the data processing agreement. For protected health information under a BAA, within 60 days, with a target of 72 hours.
-
Step 5
Learn
A blameless post-mortem within five business days, with action items tracked as work.
Register a security contact for your organisation in the portal so incident notices reach the right person rather than the account owner's inbox.
Uptime
The commitment per plan, and an honest note about what we have measured so far.
| Plan | Commitment |
|---|---|
| Free | No uptime commitment |
| Solo | No uptime commitment |
| Team | 99.9% monthly uptime |
| Enterprise | 99.95% monthly uptime |
We publish the components we monitor — control plane, gateways, cells per region, backups and a synthetic check per engine — on the status page, with incident history.
We do not publish a historical uptime figure yet. A percentage calculated over a short history is a marketing number, not a measurement, and you should not buy on it. When there is enough history for the figure to mean something, it goes on the status page.
Backups are verified, not just taken: restores are checked rather than assumed. Recovery testing on a published schedule is part of the SOC 2 work and is not yet evidenced over a full period.
Security contact and disclosure
One address, a published policy, and no legal threats for good-faith research.
Report a vulnerability to [email protected]. Machine-readable contact details are at /.well-known/security.txt. The full policy — scope, what to expect, acknowledgement within 2 business days, and fix windows by severity — is on our security page.
Privacy and data-protection questions: [email protected]. Accessibility: [email protected].
Secure hosting for protected health information is a per-instance option on Team and Enterprise under a signed Business Associate Agreement. HIPAA hosting opens after the independent penetration test. Until then you can join the waitlist: sign the BAA in the portal or contact sales.
What this page deliberately does not claim
Collected in one place so you do not have to read between the lines of the rest.
- That we hold a SOC 2 report, of either type. We do not.
- That we are “HIPAA certified”. No such certification exists; we sign a BAA and we are a Business Associate for secure instances.
- That an independent penetration test has been done. It has not been commissioned.
- That every internal network hop is encrypted. Most engine adapters are plaintext inside the cell today, restricted by network policy.
- That per-organisation customer-managed keys, dedicated hosts or flow-log retention are available. They are designed and not built.
- A historical uptime percentage. We will publish measured uptime once the status page has a meaningful history behind it, not before.