Sovraine Guard
Free Community Tier available
Your AI agents run MCP tools with no oversight. Sovraine Guard sits between your agents and your infrastructure — every action is scanned, evaluated, and audited before it executes.
Free forever. Sign up with your email to get access.
One install. Five stages. Full coverage.
From detecting your AI clients to proving every decision they make.
Detect
Finds every AI client on your machine, backs up configs, and injects governance. Zero manual setup.
Scan
Every request is scanned for PII, secrets, and regulated data on-device. Sensitive content stays local.
Evaluate
Every action passes through a multi-stage cascade. Six verdicts, per-agent risk ceilings, policy-as-code.
Decide
Allow, deny, warn, filter, escalate, or route to human. Not binary — context-aware.
Audit
Every decision logged in a tamper-evident chain. Signed, verifiable, export-ready for regulators.
Works with your AI tools
Auto-detects and protects these clients. One install covers them all.
Claude Desktop
Anthropic
Claude Code
Anthropic
Cursor
Anysphere
VS Code
Microsoft
Windsurf
Codeium
Zed
Zed Industries
Gemini CLI
Continue.dev
Continue
Get Sovraine Guard
macOS (Apple Silicon)
Enter your email to get early access. Public download at launch.
Also coming to Windows and Linux.
RUNNING AI AT SCALE?
When your team needs fleet visibility
Sovraine Guard handles the evaluation layer. Sovraine One (coming soon) adds multi-agent debate, RBAC, policy OTA, and interop connectors — for teams running AI at scale.