All AI configuration lives on one page: Third Party > AI Providers. Opening it requires the Manage AI Settings permission at Full. For an overview of the AI features themselves, see AI Features Overview.
Each supported provider — Anthropic Claude, OpenAI, and Google Gemini — has its own card with the same three fields:
********1234). Leave the field blank to keep the existing key, or tick Clear the saved API key to remove it.claude-haiku-4-5-20251001) for Anthropic Claude, gpt-4o-mini for OpenAI, and gemini-2.5-flash for Google Gemini.The provider defaults are fast, low-cost models and a good fit for the writing assistant and ticket actions; consider a larger model if report generation quality matters more than cost.
The full list of models comes from the provider itself. When the page opens, Default Model lists only the models Powercode has a built-in price for, plus the model you have saved. To see every model the provider currently offers your key, use Refresh models under the list:
The refreshed list only lasts until you leave the page. Your saved model stays in the list even if the provider stops offering it.
Spending is only calculated for models Powercode has a price for:
claude-opus-4-7,claude-sonnet-4-6,claude-haiku-4-5,claude-haiku-4-5-20251001,gpt-4o,gpt-4o-mini,gemini-2.5-flashandgemini-2.5-pro. Calls to any other model are recorded with a cost of $0, so they do not count toward the monthly caps below.
Spending on the company key is tracked per user per calendar month, priced from the provider's published per-token rates.
When a user reaches their cap, AI actions return a "Monthly AI cap reached" message until the next calendar month (or until the cap is raised). Caps apply only to the company key — usage on a personal key (BYOK, below) is logged for visibility but never counted against a cap, since it bills to the user's own account.
When a user's AI call runs on the company key, their permission group must also hold the matching per-provider permission — AI: Use Claude, AI: Use OpenAI, or AI: Use Gemini (under Modules > AI). This lets you hand out the cheap provider broadly and reserve an expensive one for specific groups. These checks do not apply to personal keys.
Each user with the Use AI features permission gets an AI Preferences section on their own User Profile page:
Personal keys are encrypted at rest, exactly like company keys.
| Symptom | Likely cause |
|---|---|
| No AI buttons appear for anyone | Enable AI features is off, or no enabled provider has a key saved. |
| No AI buttons for one user | Their group lacks Use AI features, or their preferred provider has no usable credential for them (no company-key permission and no personal key). |
| "Monthly AI cap reached" | The user hit their enterprise-key cap. Raise the default/group cap, or the user can switch to a personal key. Note a cap set to 0 blocks all usage. |
| "Provider error" responses | The API key is invalid/expired or the provider is having an outage. Check the Recent Errors & Blocked Calls list on the AI Usage Report. |
| AI features disabled message | The master switch was turned off after the page loaded, or no credential resolves for the request. |
| No API key is saved for this provider. Enter one to look up models. (Gemini: No Gemini API key is saved. Enter one to look up models.) | Refresh models was clicked with the lookup field blank and no key saved for that provider. Type a key into New API key for model lookup (optional), or save one in API Key first. |
| Unable to load models. Check the API key and try again. (Gemini: Unable to load Gemini models. Check the API key and try again.) | The provider rejected the key, or Powercode could not reach the provider. |
| No available models were returned. | The provider answered, but listed no models for that key. |