Understand API Keys in GlobalAI
API Keys in the GlobalAI platform serve as the secure gateway to your orchestrated AI resources. They allow external applications and users to consume the AI models and Agents you have configured within the platform.
Unlike the credentials you store in the Vault to authenticate outbound connections (like connecting GlobalAI to AWS), these API Keys govern inbound access. They ensure that anyone accessing your models does so within the limits and permissions you define.
Virtual keys
When you generate an API Key in the AI Orchestration module, you are creating a Virtual Key. This key acts as a proxy or abstraction layer. Instead of sharing the raw credentials of your underlying model providers with developers or client apps, you share a GlobalAI Virtual Key.
When making API calls (e.g., to the Chat Completions endpoint), the token you use for authentication is this Virtual Key.
Governance for local models
A primary function of GlobalAI API Keys is to enforce governance, particularly for Local Models hosted within your infrastructure. This prevents your internal systems from being overwhelmed by overuse. You can configure specific limits for each key:
- TPM (Tokens Per Minute): Limits the volume of text processed.
- RPM (Requests Per Minute): Limits the number of API calls allowed in a given timeframe.
- Max Budget: Sets a hard cap on the financial or resource consumption allowed.
Use cases
Departmental allocation
Create separate keys for departments such as Finance or HR, each with unique budgets (TPM/RPM) to allocate internal resources fairly and track consumption by department.
Security rotation
Set an Expiration Date (e.g., 30 days) to force periodic key rotation, ensuring that old credentials cannot be used indefinitely.
Model scoping
Restrict a key to a specific model (e.g., a fine-tuned local model) so developers cannot access other restricted models.
Related articles
Now that you understand the concepts behind API Keys, you can learn how to create them or explore how to add new models to your environment.