OpenRouter Provider for WordPress: Connectors API integration for 200+ AI models

OpenRouter Connector is a WordPress 7.0 plugin that connects OpenRouter via the Connectors API and makes it available as a provider in WP AI Client. After installation and entering an API key, any functions and plugins that use WordPress’s standard AI API can access 200+ AI models through a single provider.

What it is and why it matters

---

Architectural foundation: the WordPress AI plugin

The AI plugin (wordpress.org/plugins/ai) is an official experimental plugin from the WordPress.org team. It implements user-facing AI features on top of WP AI Client. Those features become available once OpenRouter is connected.

Current features of the AI plugin

• Title generation: a “Generate” button above the post title field that suggests options in a modal
• Excerpt generation: one-click generation from the editor sidebar
• Content summarization: generates a summary of long posts and saves it as an AI Summary block
• Alt text generation: creates descriptive alt text for images (Image block and Media Library)
• Image generation: generates images from prompts in the editor or Media Library
• Image editing: prompt-based edits (background replacement, object removal, stylization)
• Review Notes: block-by-block feedback on readability, grammar, SEO, and accessibility
• Abilities Explorer: an admin screen to browse, test, and learn registered AI abilities

Experiment Framework

All features run through the Experiments system. Each feature can be enabled or disabled independently in Settings → AI. This is opt-in: nothing runs without explicit user consent.

---

How it works: the WordPress 7.0 AI API

Connectors API

WordPress 7.0 introduced a standard mechanism for connecting external services: the Connectors API. Each connector is a registered provider with metadata (name, logo, auth method, and a link to obtain an API key).

OpenRouter Connector registers itself via wp_connectors_init:

add_action('wp_connectors_init', function(WP_Connector_Registry $registry) {
    $registry->register('openrouter', [
        'name'        => 'OpenRouter',
        'description' => 'Access 200+ AI models through a single OpenAI-compatible API.',
        'type'        => 'ai_provider',
        'logo_url'    => plugin_dir_url(__FILE__) . 'assets/openrouter-logo.svg',
        'authentication' => [
            'method'          => 'api_key',
            'credentials_url' => '<https://openrouter.ai/keys>',
            'setting_name'    => 'connectors_ai_openrouter_api_key',
        ],
        'plugin' => [
            'slug' => 'ai-connector-openrouter-wordpress',
        ],
    ]);
});

After registration, the provider appears under Settings → Connectors as a card with an API key field and a link to obtain a key at openrouter.ai.

API key source priority

WordPress 7.0 checks the API key in this order:

1. Environment variable: OPENROUTER_API_KEY
2. PHP constant: define('OPENROUTER_API_KEY', 'sk-...')
3. Database (set via the UI in Settings → Connectors)

WP AI Client: the developer entry point

Once the provider is configured, any plugin can access AI via a single fluent API:

$text = wp_ai_client_prompt('Write an SEO description for a post about WordPress 7.0.')
    ->using_temperature(0.7)
    ->generate_text();

if (is_wp_error($text)) {
    return;
}

echo wp_kses_post($text);

---

Use cases with code examples

1. Generate SEO meta descriptions

add_action('save_post', function(int $post_id, WP_Post $post) {
    if (wp_is_post_revision($post_id) || !current_user_can('edit_post', $post_id)) {
        return;
    }

    $builder = wp_ai_client_prompt(
        sprintf('Write an SEO meta description (max 160 chars) for this post: %s', $post->post_title)
    )->using_max_tokens(100)->using_temperature(0.5);

    if (!$builder->is_supported_for_text_generation()) {
        return;
    }

    $meta = $builder->generate_text();

    if (!is_wp_error($meta)) {
        update_post_meta($post_id, '_seo_meta_description', sanitize_text_field($meta));
    }
}, 10, 2);

2. REST API endpoint with AI generation

add_action('rest_api_init', function() {
    register_rest_route('my-plugin/v1', '/summarize', [
        'methods'             => 'POST',
        'callback'            => 'my_plugin_summarize_callback',
        'permission_callback' => fn() => current_user_can('edit_posts'),
        'args'                => [
            'content' => ['required' => true, 'type' => 'string'],
        ],
    ]);
});

function my_plugin_summarize_callback(WP_REST_Request $request): WP_REST_Response|WP_Error {
    $result = wp_ai_client_prompt(
        sprintf('Summarize in 3 bullet points: %s', $request->get_param('content'))
    )
    ->using_model_preference('claude-sonnet-4-6', 'gpt-5.4', 'gemini-3.1-pro-preview')
    ->using_temperature(0.3)
    ->generate_text_result();

    return rest_ensure_response($result);
}

3. Multimodal output: text + image

use WordPress\\AiClient\\Messages\\Enums\\ModalityEnum;

$result = wp_ai_client_prompt(
    'Create a chocolate cake recipe with step-by-step illustrations.'
)
->as_output_modalities(ModalityEnum::text(), ModalityEnum::image())
->generate_result();

if (!is_wp_error($result)) {
    foreach ($result->toMessage()->getParts() as $part) {
        if ($part->isText()) {
            echo wp_kses_post($part->getText());
        } elseif ($part->isFile() && $part->getFile()->isImage()) {
            echo '<img src="' . esc_url($part->getFile()->getDataUri()) . '">';
        }
    }
}

4. Capability checks before showing UI

function my_plugin_show_ai_button(): void {
    $builder = wp_ai_client_prompt('test')->using_temperature(0.7);

    if (!$builder->is_supported_for_text_generation()) {
        echo '<p>Configure an AI provider in Settings → Connectors.</p>';
        return;
    }

    echo '<button class="js-generate-ai">✨ Generate with AI</button>';
}

5. Structured JSON output

$schema = [
    'type'  => 'array',
    'items' => [
        'type'       => 'object',
        'properties' => [
            'tag'      => ['type' => 'string'],
            'relevance' => ['type' => 'number'],
        ],
        'required' => ['tag', 'relevance'],
    ],
];

$json = wp_ai_client_prompt(
    sprintf('Suggest 5 relevant tags for this post: %s', get_the_title())
)
->as_json_response($schema)
->generate_text();

if (!is_wp_error($json)) {
    $tags = json_decode($json, true);
}

---

Key features of the stack

Automatic provider discovery. Once registered in the WP AI Client registry, the connector is discovered automatically and appears in Settings → Connectors.

200+ models with one key. OpenRouter aggregates GPT, Claude, Gemini, Mistral, LLaMA, Qwen, and many other models behind an OpenAI-compatible API. Developers can set preferences via using_model_preference(), and WordPress picks the first available model.

Environment-first configuration. In production, you can set OPENROUTER_API_KEY via server config or .env so the key is not stored in the database.

Capability checks without API calls. is_supported_for_*() methods are deterministic and do not contact OpenRouter. This makes it safe to hide AI UI when the provider is not configured.

REST-friendly results. GenerativeAiResult serializes through rest_ensure_response(), so you don’t need custom response formatting.

---

Summary

OpenRouter Connector is a minimal provider plugin. It uses WordPress 7.0 standards (Connectors API + WP AI Client) to connect OpenRouter and its model catalog.

For users: one API key unlocks the AI plugin features (titles, excerpts, images, alt text, Review Notes) and model choice (Claude, GPT, Gemini, and more).

For developers: wp_ai_client_prompt() works with OpenRouter the same way it works with Anthopic, Google, or OpenAI providers. Your code stays the same; only the provider changes.

With an integrated AI Client, WordPress 7.0 turns “AI providers” into a standard connector category. OpenRouter Connector is a reference implementation of that integration.