GetInfraDesk
DocsDashboard →
Documentation
📖 Overview
Getting Started
🚀 Quick Start
🔐 AWS IAM Permissions
đŸ›Ąī¸ Security & Privacy
Decision Intelligence
✋ Decision Inbox
📄 Decision Briefs
⚡ How Approved Actions Work
â†Šī¸ Rollback Guide
Connected Signals
🔗 Signal Hub Overview
â˜ī¸ Cloudflare
🤖 OpenAI Spend
📊 Datadog
🍃 MongoDB Atlas
Support
🔧 Troubleshooting
❓ FAQ
đŸŽ¯ Pilot Program
Need help?
Email us at support@getinfradesk.com
🤖

OpenAI Spend Intelligence — Read-only beta

Reads organisation-level usage and cost metadata. InfraDesk never reads prompts, responses, threads, files, or conversations.

Required access

â€ĸ OpenAI Admin API key
â€ĸ Organisation usage access
â€ĸ Organisation cost access
Create Admin key in OpenAI dashboard →
Note: A standard API key does not have organisation usage access. You need an Admin API key.

What InfraDesk reads

  • Organisation-level token usage (input, output, cached)
  • Cost by model, project, API key
  • Number of model requests
  • Batch vs real-time usage split

What InfraDesk never reads or stores

  • Prompt bodies
  • Response bodies
  • Thread content
  • Files or vector store content
  • Fine-tuning data
  • Conversation history

Decision signals

â€ĸ AI spend spike — cost increased sharply vs previous period
â€ĸ Expensive model concentration — one model driving most spend
â€ĸ Output token bloat — output tokens 2x+ input tokens
â€ĸ Low cache utilisation — repeated prompts not cached
â€ĸ Attribution gap — spend not attributed to project or API key
â€ĸ Batch opportunity — high volume with no batch API usage