Guard get started
Run guided Guard init, connect Cloud when useful, record the first baseline, and start protecting harness launches.
Local-first runtime and approvals
Follow the pre-launch Guard loop for local evaluation, inline approvals, and blocked harness runs.
Local-first and optional cloud
See which protections stay local by default, what shared cloud features add later, and when sync is worth turning on.
Approval center and audit trail
Review queued approvals, receipt evidence, diff history, and the local approval center audit workflow.
Guard architecture
Understand the runtime layers behind harness detection, policy evaluation, receipts, and launch interception.
Harness support matrix
See how Codex, Claude Code, Cursor, Gemini, and OpenCode are handled today.
Codex harness
Set up Guard in front of Codex, understand its approval center flow, and validate project versus global MCP scope.
Claude Code harness
Use Guard with Claude hooks, workspace agents, and the strongest current deferred approval path.
Cursor harness
Keep Cursor native tool approvals while Guard handles artifact trust and runtime MCP sanity checks.
Gemini harness
Inspect Gemini extension manifests, embedded MCP servers, and the Guard approval center path for blocked changes.
OpenCode harness
Track OpenCode config, JSONC, and workspace command surfaces while preserving OpenCode native permissions.
Policy presets
Apply bundled rule sets for strict, balanced, or permissive security postures.
Mitigation guidance
Read and act on mitigation recommendations for supply-chain risks.
Testing and validation
Use the release validation loop for harness smoke coverage, manual checks, and Guard canaries.