OpenAI and Anthropic support in CloudFlow

CloudFlow introduces supports the OpenAI Administration API and Anthropic Admin API for spend guardrails, API key hygiene, and usage reporting.

As organizations adopt AI, usage often expands faster than governance. API keys multiply across teams and projects, spend is distributed across providers, and ownership becomes difficult to track. FinOps teams may have cost and usage data, but turning that data into timely action still requires manual reviews, custom scripts, and coordination with engineering teams.

This creates practical challenges: Who owns each API key? Which keys are no longer used? Is an unexpected increase in spend legitimate? What should happen when usage exceeds a team threshold? Without automation, stale credentials remain active and cost controls are applied only after overspend has occurred.

CloudFlow now supports the OpenAI Administration API and Anthropic Admin API. You can combine provider administration, cost, and usage data with filters, configurable thresholds, notifications, and automated actions in a single workflow.

This enables FinOps for AI use cases such as:

• Auditing stale or never-used API keys.
• Monitoring usage and spend across teams or workloads.
• Applying configurable spend guardrails.
• Notifying owners before intervention is required.
• Deactivating credentials that breach policy.
• Recording governance actions for reporting and review.

So, how do you put it into practice?

To make it easy to get started, we have created two tutorials explaining how to build some of these popular use-cases using CloudFlow.

Audit stale API keys

Build a read-only governance flow that inventories OpenAI and Anthropic API keys and identifies credentials that may no longer be needed. The tutorial shows how to use filters and date/time transformations to classify keys as never used or inactive for more than 90 days, then send formatted email reports to the relevant stakeholders.

Along the way, you’ll learn how to branch from provider action results, work with timestamps, filter structured data, and build table-based notifications. Follow the Audit API keys tutorial

Apply Anthropic spend guardrails

Build an automated control that evaluates Anthropic API key spend against configurable, per-key thresholds stored in CloudFlow Datastore (a simple key-value database). The flow creates a rolling reporting window, combines Anthropic cost and usage data to estimate spend attributable to each key, and identifies keys that have breached policy.

When a threshold is exceeded, the flow deactivates the key, notifies stakeholders, and records the enforcement event for future review. The tutorial demonstrates how to combine provider actions, Datastore configuration, date/time logic, custom data processing, notifications, and remediation in one workflow. Follow the Anthropic spend guardrail tutorial

Get started

Create an OpenAI or Anthropic connection, then add provider actions to your CloudFlow workflows. Please note both OpenAI and Anthropic actions are currently in early preview. Contact us to request access.

These tutorials are only a starting point — we’re excited to see what you build with CloudFlow for FinOps for AI!