BlueprintPro AI

Security

Every layer is designed to fail closed. These are the controls in place today. The full threat model lives in our docs.

Transport

TLS 1.3 with public-key pinning between the plugin and our API. No plaintext transport is exposed.

Plugin execution

The production build removes the arbitrary Python-execution endpoint entirely. Tool calls route only through a hardcoded module allowlist that is verified at plugin startup.

Model providers

Anthropic is the sole model provider. API keys are held in a managed secret store and reached via an egress-restricted gateway. Your prompts are covered by a Zero-Data-Retention contract.

Approvals

Every mutating tool call requires explicit approval in your editor unless you opt into per-run or session-scoped auto-approval.

Billing

Double-entry, append-only credit ledger, reconciled nightly against Stripe. No in-place balance mutation.

Release integrity

Plugin binaries are Authenticode-signed on Windows, notarized on macOS, and every release manifest is Ed25519-signed with keys held in an HSM.

Responsible disclosure

Report vulnerabilities to security@blueprintproai.app. We acknowledge within 24 hours.