Unified AI Cost Control

Every token counts.

Your AI spend is scattered across Claude, GPT, and Gemini — separate invoices, no attribution, no governance. Visionality gives Finance the ledger, Compliance the audit trail, and Security the controls they need — from one gateway, before the next vendor review lands on your desk.

Anthropic ClaudeOpenAI GPTGoogle GeminiAWS BedrockAzure OpenAI

No SDK migration · SOC 2 audit trail on day one · KMS-backed encryption · Deploys in 30 minutes

Enforced before the call leaves your networknot in a report the next morning
Budget limits that block — not just alertSpend Tokens are hard stops, not soft warnings
PII removed before any model sees it12 detectors, fail-closed by design
Audit trail written at the SQL layerthe app role cannot UPDATE or DELETE these rows
Governance on day one, not sprint 14deploy in 30 minutes, audit-ready on first request
Rules that actually run — not just written downmodel allowlists, PII policy, spend caps: all enforced in-flight

Claude · GPT · Gemini

Bedrock · Azure · one gateway

$80K–$2M

typical annual AI spend we govern

12

PII detectors, fail-closed

SOC 2

audit evidence in two clicks

What Visionality gives you

Four pillars of AI governance.

Most AI spend tools show you what happened. Visionality prevents what shouldn't happen — before it does.

Every call — Claude, GPT, Gemini, Bedrock, Azure OpenAI — is allocated to a project, team, and GL code the moment it lands. Spend Tokens act as hard budget envelopes that block spending before it happens. Finance exports chargeback CSVs. No more end-of-month forensics.

Cost

From AI bill to AI ledger

  • Unified ledger across Claude, GPT, Gemini, Bedrock, and Azure
  • Spend Tokens — hard budget limits per project, per model
  • Allocation rules — finance writes rules, engineers don't notice
  • Chargeback CSV export — drop straight into FP&A

Sound familiar?

This isn't a tooling gap.
It's a governance gap.

And the longer you leave it, the more it costs.

"Our AI bill tripled — Claude, GPT, Gemini across six business units. Finance can't attribute any of it."

Allocation rules map every request, across every provider, to a GL code the moment it lands. Chargeback CSVs drop straight into FP&A.

"Our compliance team is three months behind on AI vendor reviews. Every team picked their own model."

One gateway enforces model allowlists per org. Procurement reviews one vendor, not twelve.

"Legal flagged our AI feature during a SOC 2 audit. We had no audit trail whatsoever."

Five append-only audit tables — enforced at the SQL layer, not application logic. SOC 2 evidence in two clicks.

"An agent racked up $14K over a weekend. We had no circuit breaker, no visibility, nothing."

Spend Tokens are hard budget envelopes — the gateway blocks before the threshold, not after.

Three steps. Then it's running.

Deploy in 30 minutes.

01

Point your clients at the gateway

Change one environment variable — ANTHROPIC_BASE_URL, OPENAI_BASE_URL, or the Gemini endpoint — to your gateway URL. No SDK migration. No code review. The gateway speaks Anthropic Claude, OpenAI GPT, Google Gemini, AWS Bedrock, and Azure OpenAI wire formats natively. Your client code doesn't change.

02

Set your governance rules

Mint Spend Tokens for each project. Write allocation rules that map traffic to GL codes. Configure PII policy per project — block, obfuscate, or log. Takes 20 minutes for a typical setup.

03

Let finance and compliance in

Share the dashboard URL. Finance exports chargeback CSVs. Compliance browses the append-only audit log. You get the anomaly inbox so nothing surprises you at 3 AM.

Our SOC 2 auditors asked for an AI request log going back 90 days. We had it. Visionality gave us an append-only audit trail on day one — the audit committee signed off in the same meeting.

Head of Platform Engineering, 280-person healthcare SaaS·Reference available on request

Built by engineers who have shipped AI cost governance into clinical and regulated financial environments. The PII engine, append-only enforcement, and KMS encryption are ported from production systems that had to survive real audits — not proof-of-concepts.

See it before you decide.

A 30-minute demo covers your deployment, your governance setup, and your specific compliance question. No pitch deck. Just the product.

We respond within one business day. No sales sequence. No SDR handoff.

Pricing

Scales with you. Doesn't punish you for growing.

No per-seat pricing below Enterprise. Every tier includes the audit trail, PII protection, and Spend Tokens — governance isn't a paid add-on.

Starter

Infrastructure cost. Solo devs, early-stage teams evaluating AI governance.

~$7/mo

  • Full gateway — Anthropic, OpenAI, Bedrock, Azure OpenAI
  • Spend Tokens — unlimited hard budget limits
  • PII detection engine — 12 detectors, 3 modes
  • Append-only audit log
  • Anomaly detection — 4 detectors
  • Request explorer
  • 1 Clerk organization
  • Self-serve deploy — 30 minutes
Read the deploy guide

Team

Infrastructure cost. Product teams that need real cost attribution and chargeback.

~$27–60/mo

  • Everything in Starter
  • Chargeback CSV exports — month, quarter, custom range
  • Allocation rules — map traffic to GL codes
  • SaaS connector ingestion — Copilot, Cursor, AgentForce
  • Multi-org support (up to 10 orgs)
  • Per-token spend monitor with expiry
  • Anomaly inbox with severity tiers
  • Proposal review queue for agent loops
  • Email support
Request a Demo

Enterprise

SOC 2, regulated environments, healthcare, financial services.

Custom

  • Everything in Team
  • KMS-backed encryption — AWS KMS, Azure Key Vault, GCP
  • SAML/SCIM via Clerk Enterprise
  • Unlimited orgs / business units
  • Custom model allowlists per org
  • Dedicated Slack channel support
  • SOC 2 evidence package
  • SLA available
Talk to us

Starter is effectively a free tier — Neon free + Vercel Hobby + Render Starter (~$7/mo). Deploy it and use it.

Frequently asked questions

Can't find what you're looking for? Email us and someone will get back to you.

    • Do I need to pay for every seat?

      No. Starter and Team are priced by infrastructure, not by user. Everyone on your team can access the dashboard on one deployment. Enterprise has seat-based options if procurement requires it.

    • Is the audit trail really append-only?

      Yes — at the SQL layer, not in application logic. The application database role has UPDATE and DELETE revoked on the five audit tables. A deploy-time smoke check fails the rollout if that privilege was somehow restored.

    • How long does the initial deploy actually take?

      30 minutes for Starter on a cold start. 45–60 minutes if you're configuring allocation rules and PII policy at the same time. The deploy guide walks through every step.

    • What providers are supported?

      Anthropic, OpenAI, Amazon Bedrock, and Azure OpenAI. The gateway speaks each provider's wire format natively — your client code doesn't change, just the base URL.

    • Can I use my own KMS?

      Yes on Enterprise. The KeyProvider interface is designed to be swapped — AWS KMS, Azure Key Vault, or GCP Cloud KMS. Starter and Team use a master key you supply via environment variable.

    • What happens when I hit Neon or Render limits?

      Visionality uses standard managed infrastructure. If you outgrow a tier, you upgrade the underlying service. We document the upgrade path in the deploy guide.

    • Is there a free trial?

      Starter is effectively a free tier — the underlying infrastructure is either free (Neon free tier, Vercel Hobby) or very cheap (Render Starter at $7/mo). Deploy it and use it. Nothing to trial.

    • How do Spend Tokens work?

      A Spend Token is a budget envelope with a hard dollar limit. When the balance is exhausted, the gateway blocks further requests — it doesn't just alert. You can set per-project, per-team, or per-task-class limits.

    • What PII does the engine detect?

      Names, email addresses, phone numbers, SSNs, IP addresses, credit card numbers, health data (ICD codes, medication names), and several domain-specific patterns. Twelve detectors in total, tuned for low false-positive rates.