AI Operations
Context-aware AI with guardrails — not autonomous. Reads logs, configs, and metrics, then suggests scoped actions. Mutations require approval. Every AI-initiated operation is fully audited.
Supported Providers
Anthropic Claude
API key configuration, model selection
OpenAI
GPT-4 and other models
Google Gemini
Google AI integration
Microsoft Copilot
Microsoft AI integration
Ollama
Self-hosted with custom endpoint — fully air-gapped
AI Tool Capabilities
Read-Only
Auto-approved — no confirmation needed
- Browse filesystem and read files (max 1MB)
- Tail logs with regex filtering
- Get service status, list all services
- Get system performance metrics
- View active alerts
- Get incident timeline
- Compare filesystem snapshots (diff)
Mutating
Single approval required
- Write files via SafeApply workflow
- Start/stop/restart/reload services
- Create filesystem snapshots
Destructive
Double confirmation + hostname verification
- Restore from snapshots
- Execute shell commands
- System restart
Approval Workflow & Safety
Risk level indication per tool call
Diff display for file changes
Hostname confirmation for destructive operations
Full audit logging of all AI actions
Configurable: max iterations (15), max tokens (4096), temperature (0.1)
AI suggests — you decide. No autonomous execution. No black-box changes. Every action is visible, reversible, and audited.