Built for recorded
sales conversations.

Sign-in and access

• Accounts sign in through Amazon Cognito. Multi-factor authentication with an authenticator app is required for every account.
• Accounts are created by invitation only. Passwords must be at least 14 characters and mix upper- and lower-case letters, numbers and symbols.
• The web workspace and desktop app sign in through the system browser using an authorization code with PKCE. Identity-provider tokens stay on the server; the apps receive an opaque session token, and access tokens expire after 15 minutes.
• A user’s agency and role come from a membership record that is checked again on every request and every 30 seconds on live audio and event streams. Removing a member or changing their role ends access without waiting for a token to expire.

Agency isolation

• Every call, transcript, document and suggestion is stored under its agency. The agency is taken from the signed-in membership, never from anything the browser or app sends.
• Offer-level permissions decide which admins, closers and setters can see which offers, calls and documents. Closers see their own calls; agency admins can review their agency’s calls.
• Suggestions draw only on the signed-in agency’s published knowledge for the selected offer. Automated tests check that one agency cannot read another agency’s data.

Encryption

• Traffic to the workspace uses HTTPS with TLS 1.2 or 1.3; plain HTTP is redirected to HTTPS. Live audio and events use the same encrypted, authenticated connection.
• The application connects to its database over TLS and verifies the database server’s certificate.
• Database storage and its automated backups are encrypted at rest with AWS-managed keys. File storage is encrypted, blocks public access and refuses unencrypted connections.

Call audio and AI

• Magnius doesn’t store call audio. Audio is streamed to Amazon Transcribe for transcription while the closer is capturing, and only the resulting transcript is kept.
• Browser capture sends audio only. No screen frames, screenshots or video are sent or stored.
• Suggestions are generated with Anthropic Claude models through Amazon Bedrock, using transcript excerpts and your agency’s published knowledge. Offer terms in a suggestion are checked against that knowledge.
• Magnius does not train or fine-tune AI models on your calls. Shared learning across agencies is off by default, and turning the preference on does not export call data.

Infrastructure and least privilege

• The product runs on Amazon Web Services in the US East (N. Virginia) region. The database sits in private subnets and is not reachable from the internet.
• Each service has its own AWS role with only the permissions it needs, and the application uses a restricted database account rather than the administrator account.
• Services and deployment pipelines use short-lived AWS credentials; no long-lived AWS keys are stored in the services, the CI system or the desktop app. Secrets are kept in AWS Secrets Manager.
• Container images are immutable once published and scanned for known vulnerabilities on upload.

Data you control and operations

• Closers and admins can delete an ended call. Deleting it removes its transcript, suggestions, feedback and outcome.
• The database is backed up automatically every day and backups are kept for 7 days.
• Application logs are designed to exclude transcript text and access tokens. Logs are kept for 14 days.
• This website is static. It sets no cookies, uses no analytics and restricts scripts and styles to its own files with a Content-Security-Policy.

What is not in place yet

• Magnius does not hold SOC 2, ISO 27001 or any other independent certification.
• The Windows installer is not yet code-signed, so Windows may show an unknown-publisher warning. The macOS app is in early access and is not yet notarized by Apple.

Report a security issue

Email [email protected] with “Security report” in the subject. Please include the steps to reproduce and do not access other customers’ data. See the subprocessor list and our data principles for how data flows.