The Manage LLM Providers screen lets you view, add, edit, and remove the AI providers that power your assistants. Open it from Agent Builder → Settings → Providers tab.
Agent Builder ships with nine built-in providers — OpenAI, Anthropic, xAI (Grok), Google Gemini, Mistral, Meta (Llama), Cohere, Ollama, and Agentic — but you can register any OpenAI-compatible endpoint as a custom provider.
Provider Table
The main view is a sortable table listing every registered provider:
| Column | Description |
|---|---|
| Name | Display name shown in the provider dropdown throughout the plugin. Built-in providers show their brand icon next to the name. |
| Slug | Unique machine identifier (e.g. openai, anthropic). Used internally for API routing and configuration storage. |
| Endpoint | The API endpoint URL. May include placeholders like %MODEL% and %KEY% that are substituted at runtime. |
| Default Model | The model used when no specific model is selected. For example, gpt-4o for OpenAI or claude-sonnet-4-20250514 for Anthropic. |
| Auth Type | How the plugin authenticates with the provider’s API (see Authentication Types below). |
| Format | The request format used to communicate with the API (see Request & Response Formats). |
| Actions | Edit, Set Default, and Delete (custom providers only — built-in providers cannot be deleted). |
The currently active default provider is highlighted and marked with a “Default” badge. Click Set Default on any row to change it.
Adding a Custom Provider
Click the Add New Provider button to open the provider form. Fill in the following fields:
Provider Name
A human-readable label displayed throughout the plugin interface. Choose something descriptive — for example, “My Local LLM” or “Company API Gateway”.
Icon URL
Optional. A URL to a small icon image (recommended 24×24px) displayed alongside the provider name. Leave blank to use a generic icon.
Slug
A unique machine identifier using only lowercase letters, numbers, and hyphens. This cannot be changed after creation. Examples: my-local-llm, company-gateway.
Endpoint URL
The full API endpoint URL. You can use these placeholders:
%MODEL%— replaced with the selected model name at runtime%KEY%— replaced with your API key (used with the “Key in URL” auth type)%OLLAMA_URL%— replaced with the configured Ollama base URL
Example: https://api.example.com/v1/chat/completions
Default Model
The model identifier used when the user hasn’t specified one. This should match the model name expected by the provider’s API (e.g. gpt-4o, claude-sonnet-4-20250514).
Available Models
A list of all models supported by this provider, one per line. These populate the model dropdown when this provider is selected in Settings. You can type them manually or use the Refresh Models button to auto-fetch the list from the provider’s API.
Authentication Type
How the plugin sends credentials to the provider:
| Type | Description | Used By |
|---|---|---|
| Bearer Token | Sends the API key as Authorization: Bearer <key> in the request header. The most common authentication method. |
OpenAI, xAI, Mistral, Meta, Cohere |
| Anthropic | Sends the API key via the x-api-key header, as required by Anthropic’s API. |
Anthropic |
| Key in URL | Embeds the API key directly in the endpoint URL using the %KEY% placeholder. |
Google Gemini |
| None | No authentication required. Suitable for local models. | Ollama |
| Built-in Constant | Uses the AGENTIC_AI_API_KEY PHP constant defined in your WordPress configuration. Used by the built-in Agentic provider. |
Agentic |
Request Format
Determines how the plugin structures outgoing API requests:
- OpenAI — Standard OpenAI chat completions format. Compatible with most providers and OpenAI-compatible APIs.
- Anthropic — Anthropic Messages API format with system prompts sent separately.
- Google — Google Gemini
generateContentformat. - Agentic — Custom format used by the built-in Agentic provider.
Response Format
Determines how the plugin parses incoming API responses:
- OpenAI — Standard OpenAI response structure (
choices[0].message.content). Works with most OpenAI-compatible providers. - Cohere — Cohere Command response format.
- Ollama — Ollama’s native response format (used internally by the Agentic Ollama bridge).
Get Key URL
Optional. A link to the provider’s API key management page. Displayed as a convenience link in the General settings tab when this provider is selected.
Editing a Provider
Click Edit on any provider row to modify its settings. All fields can be updated except the slug of built-in providers. Changes take effect immediately after saving.
Setting the Default Provider
Click Set Default on any provider row to make it the active provider. The default provider is used for all AI assistant interactions unless overridden at the agent level. The change is applied immediately — no additional save step is required.
Refreshing Available Models
When editing or adding a provider, click the Refresh Models button next to the Available Models field. The plugin will query the provider’s API to retrieve the current list of available models and populate the field automatically. This requires a valid API key to be configured in your General settings.
If the refresh fails (due to an invalid key, network issues, or an unsupported provider API), you can enter model names manually — one per line.
Deleting a Provider
Click Delete on a custom provider row, then confirm the action. Deleted providers are permanently removed.
Note: Built-in providers (OpenAI, Anthropic, xAI, Google, Mistral, Ollama, Meta, Cohere, and Agentic) cannot be deleted. The Delete button is only available for custom providers you have added.
Built-in Providers
Agent Builder includes nine pre-configured providers ready to use:
| Provider | Default Model | Auth Type | Notes |
|---|---|---|---|
| OpenAI | gpt-4o | Bearer Token | Most popular. Supports GPT-4o, GPT-4o-mini, o1, o3-mini, and more. |
| Anthropic | claude-sonnet-4-20250514 | Anthropic (x-api-key) | Claude models. Uses a dedicated authentication header. |
| xAI (Grok) | grok-2-latest | Bearer Token | Grok models from xAI. |
| Google Gemini | gemini-2.0-flash | Key in URL | API key embedded in the endpoint URL. |
| Mistral | mistral-small-latest | Bearer Token | European AI provider with efficient models. |
| Meta (Llama) | Llama-4-Scout-17B-16E-Instruct | Bearer Token | Open-weight Llama models via Meta’s hosted API. |
| Cohere | command-r-plus | Bearer Token | Enterprise-focused models with RAG capabilities. |
| Ollama | llama3.2 | None | Run models locally. No API key needed — requires Ollama installed on your server. |
| Agentic | agentic | Built-in Constant | Built-in provider using the AGENTIC_AI_API_KEY constant. |
Tips
- You can use any OpenAI-compatible API endpoint (such as OpenRouter, Together AI, or a local LM Studio server) by adding a custom provider with the OpenAI request and response format.
- To connect your API key, go to Agent Builder → Settings → General and enter it in the API Key field. The key is used by whichever provider is set as the default.
- If you switch the default provider, make sure a valid API key for that provider is configured.
- For local models via Ollama, set the Ollama URL in General settings — no API key is required.
Related Documentation
- Connecting to an LLM Provider — Initial API key setup and first connection
- Choosing an LLM Provider — Comparison guide to help you pick the right provider
- Settings — Overview of all Agent Builder settings