Security
How we look after your data
How Muxila is built to handle your data, in plain terms. We don't hold security certifications yet, and we'll say so here when that changes.
Isolation
- A separate Postgres database, file storage and encryption key for every company
- Sign-ins and user accounts live in your company's own database
- MCP access uses each person's own sign-in, never a shared key
Encryption
- TLS for all traffic
- Connector tokens and API keys encrypted with a key per company (envelope encryption)
- Daily encrypted backups, kept for 30 days
AI models
- Your data is not used to train AI models
- We use API providers that don't train on API data by default
- Personal details are masked in our logs and AI traces
- Answers come from your documents, with a source for every claim
Access
- Permissions by group, role and person, with source-level sharing respected
- Sign in with Google or Microsoft, SAML on request
- Workflow actions run with the person's own access, never a shared admin key
- Audit log of admin actions
Your data, your call
- Export everything at any time
- Conversation retention you can shorten per account
- Delete on request, and full deletion 30 days after you leave
- Data processing agreement (DPA) on request
Hosting
- Each company chooses EU or US hosting
- Infrastructure runs on Railway and Cloudflare
- Separate staging and production environments
Sub-processors
Services that process data for us
We'll update this list before adding a new sub-processor.
| Railway | Application hosting and databases |
| Cloudflare | File storage (R2), DNS and bot protection |
| OpenAI, Anthropic, Google | AI model inference, depending on the feature |
| Resend | Transactional email |
| Sentry | Error monitoring |
Found a security issue?
Email contact@muxila.com. Put "Security" in the subject. We'll reply and keep you updated until it's fixed.
Questions about your data?
Ask us for a DPA or anything else your security review needs.