Current security and control model
This page describes product controls visible in the current Relvora implementation. It is not a certification statement and does not claim compliance with a specific security framework unless expressly stated.
Workspace isolation
Application data is associated with a workspace, and row-level access policies are used to restrict authenticated users to workspaces they are members of.
Integration credentials
Supported provider credentials are handled server-side and sensitive integration credentials are stored in encrypted form within the application architecture.
Role based access
Workspace roles and granular permissions control access to conversations, Agents, integrations, notifications, billing, and administrative functions.
Draft-first Agent changes
Editing a published Agent creates an isolated draft. The live Agent continues using the published configuration until a user explicitly publishes a new version.
Human controls
Authorized users can pause, resume, or remove an Agent from a conversation and can move conversations between active, done, and deleted states.
Guardrails and escalation
Agents support explicit guardrails and are designed to escalate when important information is missing, a human is requested, or the appropriate response is uncertain.
Infrastructure
Relvora currently uses managed infrastructure and service providers including Vercel and Supabase, with OpenAI used for AI-powered functionality and connected third-party providers used when you enable those integrations. Security responsibilities are shared between Relvora, those providers, workspace administrators, and end users.
Customer responsibilities
Protect your account credentials and provider API keys, assign the minimum necessary permissions, remove users who no longer need access, review Agent instructions before publishing, use test mode for meaningful changes, and avoid placing secrets or unnecessary sensitive information into Agent prompts or connected conversations.
Report a security issue
If you believe you found a security issue affecting Relvora, contact hello@hirschventuresgroup.com with enough information for us to reproduce and investigate the issue. Do not publicly disclose an unresolved issue that could expose customer data or systems.