Built for teams whose compliance reviewer
reads every commit.
SysNav earns its place in regulated environments by leaving your keys where they belong — with you.
How SysNav keeps secrets, secret.
Local-first by design
SSH keys, credentials, and env vars never leave your device. The shell runs on your machine; the AI orchestrator sees only redacted, ephemeral context.
Minimal retention
Your prompt and command context are processed in memory to answer the request — not retained as a SysNav dataset. We keep only a metadata audit log (command text and approvals), never file contents or secrets.
Redaction pipeline
An on-device scrubber strips common key formats (AWS, GCP, JWT, private-key blocks, credential env vars, etc.) before any token is sent upstream.
BYO LLM
Today SysNav runs on Anthropic Claude with an OpenAI fallback. Bring-your-own-key and additional providers (Azure, Bedrock, self-hosted Llama/Mistral) are on the roadmap.
Tamper-evident audit
Every command, approval, and AI context is recorded in an append-only audit log with a SHA-256 hash chain, with an optional mirror to your own S3 bucket. Exportable as JSON.
Safety gate
Destructive commands are blocked in Ask mode and require explicit approval in Agent mode, enforced by an allowlist-first command gate with deny-patterns. Broader policy controls (time-windows, region scoping) are on the roadmap.
Frameworks, status, reality.
SysNav is in public beta. None of the frameworks below are certified yet — this table reflects our roadmap and current data-handling practices, not completed audits.