System Design

Data VPN Firewall Overview

Traditional VPNs mask your network identity (IP address) while leaving raw prompt payloads untouched. Spendplane is a Data VPN Firewall for AI: it masks your payload identity—scrambling sensitive data points, PII, and API secrets before egress across Mobile, Desktop, and Production APIs.

How do I route v0 traffic to Llama-3?

Why “Data VPN Firewall”?

Traditional VPNs hide where your traffic comes from by masking your IP address—but they send 100% of your raw prompt text straight to the AI provider. Spendplane routes your AI traffic through a dedicated Data VPN Firewall: instead of masking your IP, it inspects and scrambles sensitive data points, API keys, and PII inside the tunnel before anything reaches an external model.

Two-Mode Architecture: On Mobile and Desktop (AI Safe), Spendplane uses a lightweight on-device local VPN/proxy tunnel to inspect and scramble prompts before they leave your device. For backend applications (Spendplane Gateway), it acts as an application-layer pre-egress firewall via a single BASE_URL swap—with optional self-hosted VPC deployment for Enterprise.

Data Flow: Client Surface to Data VPN Firewall

MOBILE / DESKTOP / API
DATA VPN FIREWALL
APPROVED MODEL / VPC

Application-Layer Gateway & On-Device Tunnel

Spendplane operates as an HTTP reverse proxy gateway with tenant-scoped API keys and TLS 1.3 transport. Optional policy rules perform pre-egress PII redaction in memory via the Policy Inspector before requests reach your selected model provider.

Governance & Compliance

Every request emits structured metadata telemetry (timestamp, model, token counts, latency, and redacted entity types—never raw prompt bodies). The Spendplane Gateway enforces model allowlists, region routing, and workspace spend caps.

Gateway Transport & Policy Status

Transit
TLS 1.3 / mTLS
Auth
Workspace Key (sp_live_...)
Shield
PII Redaction: Active
Trace Hub: Manual v1.0.4 / Gateway Architecture: Spendplane Gateway