Settings
Configure agent model tiering, tool policies, and privileged access.
Model Tiering
Model assignment per agent role. This is determined by the backend and is read-only in this view.
| Agent Role | Model Tier |
|---|---|
| Aegis | Haiku |
| Keeper | Sonnet |
| Scribe | Sonnet |
| Indexer | Haiku |
| Auditor | Haiku |
| Sentry | Haiku |
Tool Policies
Access control rules that govern which agents can invoke specific tools. Displayed here for reference.
Messaging access
Only Aegis holds Telegram/messaging access for external communication.
Vault staging write
Only Keeper and Auditor can write to vault staging. All other agents are read-only.
Shell execution
OperatorX is the only agent with direct shell execution capability. All other agents are sandboxed.
Vault read access
All agents have read access to vault content for context retrieval.
Configuration changes
Only OperatorX can modify agent and system configuration outside the vault.
OperatorX
Privileged troubleshooting agent with direct command execution access on your Pi. Enable only when needed.
OperatorX is disabled
No privileged agents are currently active.