Trust center
Enterprise-Grade Security
Encryption, SSO, customer isolation and data retention controls for every request.
Encryption in transit and at rest
Secrets get an additional layer of encryption.
Encryption in transit
All traffic to the service is encrypted using industry-standard protocols.
Encryption at rest
Stored data is encrypted at rest using industry-standard encryption.
Protected secrets
Provider keys, API tokens and other credentials are encrypted before they are stored and are never shown again after you add them.
Managed key storage
Encryption keys are kept in a managed secret store, separate from the data they protect, and can be rotated.
SSO, SCIM and least-privilege access
People sign in through your identity provider. Keys and workloads get only the access they need.
Single sign-on
Connect your identity provider, verify your domain and require single sign-on for everyone in your company.
SCIM provisioning
Users and groups are provisioned and removed automatically from your identity provider.
Roles and organizations
Admins manage people, keys and policies for each organization, and admin rights in one organization do not carry into another.
Scoped API keys
Each key carries only the permissions and routers it needs, can expire, and can be revoked at any time.
Workload identity
Build systems and services can sign in with their own identity instead of storing long-lived keys.
Multi-factor authentication
Enforced through your SSO provider.
Customer isolation
Every router, key, prompt, budget and traffic log belongs to one organization.
Separate by design
Configuration and request records are scoped to the organization that owns them, and access is checked on every request.
Policies for AI agents
Policies control which agents people can run, which models they can use and how much they can spend.
Provider keys never leave the service
Applications hold one Multi-Router key. Your provider keys are never returned.
Only the providers you choose
Requests go only to the model providers you route to, using your own provider accounts.
Provider keys stay private
Provider keys are never returned by the dashboard or the API after you add them.
No training on your content
Your prompts, responses and files are not used to train models.
Redaction before it leaves
Guardrails can remove sensitive data from requests before they reach a provider, and block answers that fail a check.
Data retention controls
Choose what is stored for each request and how long it's kept.
You choose what is stored
Store full requests, metadata only, anonymized usage, errors only, or nothing at all.
Content only when you choose
Prompts and responses are stored only when you turn that on.
Limited retention
Stored request data expires automatically after a fixed period in the hosted service.
Deletion
When an account is closed, its data is deleted within 30 days.
Hosted on AWS in the United States
Operated with least-privilege access.
Hosted on AWS
The hosted service runs in AWS data centers in the United States.
Network isolation
Application services are not directly reachable from the internet.
Tested changes
Every change is tested in a staging environment before it reaches production, and is rolled out without downtime.
Least-privilege access
Engineers use short-lived, limited credentials to reach production systems.
Self-host for full data sovereignty
Keep all data inside your own cloud account.
Your cloud, your data
Configuration, keys, request records and stored requests stay in stores you own.
Nothing else leaves
Only calls to the model providers you configure, and optional catalog updates, leave your network.
Data protection and compliance reviews
Subprocessors
Third parties that process data to run the hosted service.
| Subprocessor | Purpose | Location |
|---|---|---|
| Amazon Web Services | Cloud hosting and infrastructure | United States |
| Stripe | Payments and invoicing | United States |
| Formspree | Receiving contact and sales forms | United States |
| Google and GitHub | Sign-in, when you choose it | United States |
Report an issue
If you believe you have found a vulnerability, email security@multi-router.ai with the details and steps to reproduce. We acknowledge reports within 2 business days and ask that you give us reasonable time to fix the issue before disclosing it.
Questions from security reviewers
The questions security teams ask most often, answered.
Data handling
What data does Multi-Router store?
Account details, the configuration you create, your encrypted provider keys, and a record of each request. What each request record contains depends on the storage setting you choose.
Do you store prompts and responses?
Only if you turn that on. Every other storage setting keeps prompts and responses out of storage.
How long is data kept?
Stored request data expires automatically after a fixed period in the hosted service. Configuration is kept while your account is active.
How is data deleted when we leave?
When an account is closed, you can export your data for 30 days. After that it is deleted within 30 days.
Do you train models on our data?
No. Your prompts, responses and files are not used to train models.
Encryption
Is data encrypted in transit and at rest?
Yes. All traffic is encrypted in transit, and stored data is encrypted at rest using industry-standard encryption.
How are secrets protected?
Provider keys, API tokens and other credentials receive an additional layer of encryption before they are stored, with keys held in a managed secret store.
How are passwords stored?
Passwords are hashed with an industry-standard password hashing algorithm and are never stored in plain text.
Identity and access
Do you support single sign-on?
Yes. You can connect your identity provider, verify your domain and require single sign-on for everyone in your company.
Do you support SCIM?
Yes. Users and groups are provisioned and removed automatically from your identity provider.
Do you support MFA?
Multi-factor authentication is enforced through your SSO provider.
How is API access controlled?
Each API key carries only the permissions and routers it needs, can expire and can be revoked at any time. Build systems and services can sign in with their own identity instead of stored keys.
How are customers separated from each other?
Every resource belongs to one organization, and access is checked on every request. Admin rights in one organization do not carry into another.
Model providers
Where do our requests go?
Only to the model providers you route to, using your own provider accounts.
Can anyone read our provider keys?
Provider keys are encrypted when stored and are never returned by the dashboard or the API after you add them.
Can sensitive data be removed before it reaches a provider?
Yes. Guardrails can remove sensitive data from requests and block answers that fail a check.
Infrastructure and operations
Where is the service hosted?
On AWS in the United States.
Can we host it ourselves?
Yes. A self-hosted deployment keeps all data in your own cloud account. Only calls to your model providers, and optional catalog updates, leave your network.
How are changes deployed?
Every change is tested in a staging environment first and rolled out to production without downtime.
How is production access controlled?
Engineers use short-lived, limited credentials to reach production systems.
How do you handle incidents?
We investigate and contain incidents as soon as they are detected. If customer data is affected, we notify affected customers without undue delay and within 72 hours of becoming aware of it, as set out in our Data Processing Agreement.
Need more for your review?
We complete security questionnaires and share our security documentation under NDA.