Skip to content
Power Prompt Documentation

System API Keys and Authorised AI Models

Power Prompt maintains a strict separation between two categories of inference credentials: organisation/team keys (used for user testing) and System API Keys (managed centrally to power native platform features).

1. The Purpose of System API Keys

System API Keys represent the shared infrastructure foundation of Power Prompt. They power built-in intelligent features transparently for all users:

  1. The “Improve” feature in prompt editing: Instantly analyses, refines, and formats draft prompts directly in the editor in a single click.
  2. The “Prompt Architect” feature in the Prompt Lab: Used for creating and structuring complete, production-ready prompts from natural language requirements.
  3. The “Prompt Improver” feature in the Prompt Lab: Used for analysing, restructuring, clarifying, and strengthening existing prompts with advanced engineering controls.

By centralising these system-level credentials at the platform level, the user experience remains smooth and operational from day one, without requiring individual user API keys.

2. Authorised AI Models Whitelist

To ensure regulatory compliance, data residency adherence, and cost governance, administrators maintain the enterprise whitelist of authorised AI models:

Model Whitelisting Criteria:

  • Data Processing Residency: Confirming regional boundaries (e.g., European Union processing for GDPR compliance).
  • Vendor Privacy Guarantees: Verification of zero-retention policies and formal commitments that customer inputs are never used for model training.
  • Security & Encryption Standards: TLS 1.3 encryption in transit and SOC 2 Type II vendor compliance.
  • Cost-to-Capability Alignment: Approving optimal model tiers for specific departmental workloads.

3. Configuration and Key Rotation Policy

  1. In the platform administration console, navigate to System Keys & Models.
  2. Enter enterprise API credentials for approved providers (OpenAI, Anthropic, Google).
  3. Toggle individual models on or off according to corporate security directives.
  4. Establish system-wide rate limits and monthly infrastructure spending caps.
  5. Enforce regular credential rotation (recommended every 90 days) to adhere to enterprise cybersecurity standards.