Quickstart

Prototype to Production
in 60 seconds.

Point your AI client or SDK to the Spendplane Gateway (`https://api.spendplane.com/v1/openai`), apply workspace budget caps and pre-egress PII redaction, and inspect policy decisions in real time below.

01

Set Base URL

Point your OpenAI or Anthropic SDK to https://api.spendplane.com/v1/openai or /v1/anthropic.

02

Authorize Workspace Key

Add your sp_live_... key in Cursor, Bolt, v0, or your backend environment variables.

03

Verify Redaction & Routing

Test pre-egress scrubbing in the Policy Inspector below and review metadata traces in the Control Plane.

How do I route v0 traffic to Llama-3?

Interactive Policy Inspector

Test pre-egress redaction rules directly in the browser. Paste a sample prompt below to inspect how emails, phone numbers, and API credentials are tokenized before upstream provider egress.

Policy Inspector

Pre-Egress PII Scanner

Live Logic Active
Low Latency
Hey, my email is [EMAIL_1] and my development key is [API_KEY_1]. Please check the logs at [IP_ADDRESS_1].

Governance Shield Active: Your keys and PII are stripped before they leave your control boundary. LLMs never see your raw identity.

The governance tunnel

Control exactly what leaves your network.

Spendplane captures the request, applies policy, picks the cheapest viable lane, and shows the live decision instead of hiding it behind the agent.

The Scrubbing Layer

Sub-5ms pre-egress redaction tokenizes PII and API secrets ([EMAIL_1], [SECRET_1]) before prompts leave your boundary, while enforcing hard workspace spend caps with zero prompt retention.

  • Sub-5ms automatic PII & secret redaction
  • Zero prompt body storage
  • Hard workspace spend caps
Original payload: draft reply for John Doe (jdoe@acme.com)...
Local Fast Lane (Selected)
Free
Budget Cloud
Optimal
Premium Assist
$0.04/req

Intelligent Routing

Not every prompt needs a frontier reasoning model. Spendplane dynamically routes your payload to the cheapest capable model based on task complexity, slashing your cloud bill without sacrificing quality.

  • Model cascading capabilities
  • Cost-threshold fallbacks
  • Latency-aware selection

Trace, Restore & Deliver

Watch the full lifecycle in real-time. Once the LLM responds with scrubbed tokens, Spendplane securely re-maps them to their original values and delivers a complete, private response to your app — zero PII ever leaked.

  • Metadata-only audit logging
  • Optional token re-insertion on return
  • Cost attribution per request
  • Automated pre-egress PII & secret redaction
Spendplane terminal

_

Spendplane Recipes

Developers in 2026 love "copy-pasteable" success. Grab these pre-validated configs to kickstart your production stack.

Cost Optimization

The $5/mo Bolt.new Portfolio

Routing all non-critical architectural tasks to Llama-3.1 via local proxy while keeping frontend-hot-reloading in the cloud.

.env
SPENDPLANE_ROUTING=hybrid
LOCAL_GPU_PRIORITY=high
MAX_CLOUD_SPEND=5.00
config.json
{ "models": ["llama-3.1-8b"], "provider": "spendplane-local" }
Governance

The HIPAA Healthcare Bot

PII Scrubbing + 100% Data Retention at Source. Ensure zero patient data ever hits third-party model providers.

.env
SPENDPLANE_ENFORCER=strict
PII_REDACTION=enabled
RETENTION_POLICY=source_only
config.json
{ "security": { "pii_scanner": "deep", "vault_key": "VPC_ROOT" } }

Engineering Trust

Spendplane is built for controlled AI adoption with zero prompt body storage, AES-256-GCM encrypted key vaulting, and optional on-device inspection via Spendplane AI Safe.

Security Whitepaper

Architecture & control details.

Zero Prompt Storage

Transient memory-only transit.

On-Device AI Safe

Local inspection with zero cloud proxy.

Region-Aware Routing

US and EU gateway routing controls.