Trust
AI should be useful, controlled, and safe
ElixirAgent is built around tenant isolation, backend-only secret storage, role control, and a clear rule: when a human takes over, AI stops sending messages.
Ownership rule
One owner per conversation at a time
AI replies automatically only until an operator takes the conversation. After handoff, AI can help with drafts but does not send them without a human.
Trust
Reliability without heavy explanations
Channels, roles, and AI modes stay transparent: who replies, what is connected, and when a human is needed.
Tenant isolation
Each company stays inside its own scope.
Secrets in backend
Tokens and credentials stay off the browser.
Human priority
After handoff, AI stops sending on its own.
Tenant isolation
Each company has its own workspace, members, settings, and channel connections.
Encrypted secrets
Channel tokens and credentials are stored only on the backend in encrypted form.
Session safety
Access is handled through secure HTTP-only cookies, refresh flow, and logout/revocation.
Audit trail
Connection changes, tests, and critical admin actions are recorded for accountability.
AI boundaries
The model uses company knowledge and escalation rules instead of inventing unsupported answers.
Operator ownership
When a manager takes a chat, automatic AI sending stops until the mode is changed back.
For the client