| Key features | - Root cause proven by read-only checks against live systems
- One-tap human approval in Slack, Symphony, PagerDuty, phone or email
- Shadow mode by default, per-service autonomy with two-person promotion
- Tamper-evident, hash-chained, signed audit ledger
- On-prem and air-gapped deployment with no inbound ports
- Connectors for Prometheus, Grafana, Splunk, Datadog, Git hosts, Jira, ServiceNow, Kubernetes and SQL databases
- Dry run before execution, every time
- SIEM-ready: syslog to Splunk, QRadar or any SIEM
- Secrets never stored: redacted before anything is written
| - TUN interception for system-wide proxy routing without environment variables
- Split traffic by domain, IP address, and application
- Multiple API configuration profiles with one-click switching
- Local token usage and cost estimation by date, model, and session
- Session analysis with context length warnings and cost breakdown
- Support for VLESS, VMess, Trojan, Hysteria2, AnyTLS, Shadowsocks, and Base64 subscriptions
- Automatic detection of claude and codex process traffic
- DeepSeek Harness usage tracking
- Configuration file auto-write for Claude Code settings.json and Codex config.toml
|
|---|
| What makes it different | - Proves root causes through live system checks, not guesses
- Operates entirely on-premises with no inbound connections
- Measured median time to fix from alert to verified resolution
- Every action is auditable, hash-chained and signed
| - All three functions—proxy routing, API management, and usage tracking—integrated into one Mac app designed for Claude Code and Codex developers
- Local cost estimation based on actual token usage read from local logs, no forwarding gateway required
- Support for switching between official and custom relay APIs without environment variable changes
|