Product overview
Built for the complexity of enterprise AI
Keystone isn’t just a dashboard; it’s an intelligence layer that sits between your engineering architecture and your finance team.
Multi-provider intake
Current integrations cover OpenAI, Anthropic, Google Cloud billing exports for Vertex and Gemini spend, AWS Bedrock, and bounded CSV imports. Keystone normalizes the supported usage and cost records into one finance-oriented view.
Keystone Units (KU)
1 KU = the standard cost of a GPT-4-class AI task. Think kilowatt-hour for AI: a single normalized unit so you can compare OpenAI to Anthropic to Bedrock on equal terms.
Real-time anomaly detection
Did a dev leave a script running? Our engine catches spend spikes across your endpoints in minutes, not at the end of the billing cycle.
Finance review queue
Review spend anomalies and budget-risk findings with explicit owners, notes, and resolution status. Finance-side AI discovery remains feature-flagged until its data migration is enabled.
Managed deployment
Keystone currently runs on managed application and PostgreSQL infrastructure with tenant-scoped access controls. Self-hosted deployment is not part of the current release.
AES-256 encryption
Your API keys are encrypted at rest using AES-256. We never log payload data—only token counts and metadata.
The "Three Views" framework
Managing AI isn't like managing SaaS. The usage is variable, the pricing changes weekly, and the ROI is difficult to quantify. We built the "Three Views" framework to give finance leaders concrete tools.
- Rearview: What happened and why.
- Windshield: What will happen next month.
- Blind spots: What is quietly burning cash right now.