Security
How Meddle protects agent settings, call data, and the credentials your agents use.
Platform boundaries
Meddle is multi-tenant. The tenancy unit is the workspace, and every agent belongs to exactly one workspace.
How isolation is enforced
Every request to agent data is checked against the same two questions:
- Is the requesting user a member of the workspace that owns this agent?
- Or does an explicit share grant them access to this specific agent?
Writes also require that share to carry edit permission. The same check applies uniformly to every access path: the web application, the MCP server, and any service that authenticates as a user.
Continuous verification
Tenant-isolation rules are verified automatically before every release. Automated secret scanning, static analysis and dependency checks run on every change before it ships.
Authentication
| Method | Status |
|---|---|
| Email and password | Available |
| Google sign-in | Available |
| Time-based one-time password, second factor | Available, per user |
| Recovery codes for second-factor loss | Available, stored hashed |
Access control
Workspace membership carries a role. Roles govern what the application permits (inviting members, managing billing, changing workspace settings). To scope access to a single agent, use per-agent collaborators.
See Access control for roles, collaborators and the audit log.
Credentials
Secret variable values are held in an encrypted vault. API keys and publishable keys are stored as hashes.
See Credential handling for the vault-only dialog path.
Call recordings
Recordings are held in a private storage bucket, gated by the same agent-access predicate that governs the rest of the platform.
Operators retrieve a recording through a short-lived signed link generated at the moment of the request.
While a call is running
Caller speech is treated as data, not instruction, and outbound requests are constrained to prevent agents reaching internal infrastructure.
See Runtime safeguards.
Reporting a vulnerability
Send details to hello@verticalai.com.au with "security" in the subject line.
We will acknowledge receipt and keep you informed while we investigate.