Providers are managed on Kriti AI → Integrations Hub. The page is grouped by what each provider can create: text, image, audio and video.

Note: In Kriti AI 1.0.x this menu item was called MCP Server.

• To turn a provider on, paste its API key (or, for Ollama and Local AI, its server URL) and click Save Settings. There is no separate on/off switch.
• To turn a provider off, clear the field and save.

You can turn on as many providers as you like, for example OpenAI for articles and Stability AI for images.

Every card has a Test connection button. It sends a read-only request to the provider, such as listing its models, and shows a green Connected tick or the provider's own error message.

The test never generates content, never uses credits and does not save the key. Click Save Settings afterwards.

Content type	Providers
Text (articles)	OpenAI, Anthropic Claude, Google Gemini, DeepSeek, Groq, OpenRouter, Ollama
Image	OpenAI (DALL·E), Google Gemini, Stability AI, Replicate, Local AI
Audio (text to speech)	OpenAI TTS, ElevenLabs, Google Gemini TTS, Local AI
Video	OpenAI Sora, Google Veo, Local AI

Note: Anthropic, Groq, OpenRouter, Stability AI, Replicate, ElevenLabs and Local AI were added in Kriti AI 1.1. Update the plugin if you don't see them.

• OpenAI (text, image, audio, video): platform.openai.com/api-keys
• Anthropic Claude: console.anthropic.com
• Google Gemini (text, image, audio, video): aistudio.google.com/apikey
• DeepSeek: platform.deepseek.com/api_keys
• Groq: console.groq.com/keys
• OpenRouter: openrouter.ai/keys
• Stability AI: platform.stability.ai
• Replicate: replicate.com/account/api-tokens
• ElevenLabs: elevenlabs.io

OpenAI and Gemini have a separate card for each content type, so you can use a different key for images, audio and video.

Cloud providers use their recommended default model, so you don't have to pick one to get started. To use a different model for a single generation, choose it from Model in AI Studio.

OpenRouter uses openrouter/auto by default, which routes to a suitable model. You can enter any OpenRouter model ID in AI Studio.

Run open models such as Llama, Qwen or Mistral on your own computer or server with Ollama. Nothing is sent to a third party.

1. Install Ollama and pull a model, for example ollama pull qwen3.5.
2. On the Ollama card, enter the Server URL, usually http://localhost:11434.
3. In Model, enter the full model name including its tag, for example qwen3.5:latest.
4. Click Test connection, then Save Settings.

Note: Your WordPress server must be able to reach the Ollama URL. localhost only works when Ollama runs on the same machine as WordPress.

Local AI connects to any server that offers the OpenAI-compatible /v1 endpoints, such as LocalAI.

1. On the Local AI (Image), (Audio) or (Video) card, enter the server's base URL in API URL, for example http://localhost:8080.
2. Enter the Model name your server serves.
3. Add an API key only if your server requires one.

The Mock provider returns placeholder text, images, audio and video, so you can try the whole workflow (generate, queue, review, publish) without an account anywhere.

It is offered only when no other provider is set up for that content type, and steps aside as soon as you add one.