=== Agent Builder ===
Contributors: agenticplugin
Tags: ai, artificial intelligence, automation, chatgpt, claude
Requires at least: 6.4
Tested up to: 6.9
Requires PHP: 8.1
Stable tag: 2.9.45
Donate link: https://agentic-plugin.com/donate/
License: GPL-3.0-or-later
License URI: https://www.gnu.org/licenses/gpl-3.0.html

Use AI to assist with Content, E-Commerce, SEO, CRM, Security, Maintenance and more - with full approval workflows and data security.

== Description ==

[Agent Builder](https://agentic-plugin.com/) lets you create AI agents using plain job descriptions that turn Wordpress into an intelligent, autonomous, more secure and well managed CMS - deploy agents to write content, optimize SEO, monitor security, generate images and videos, normalise your data and generate useful reports.

= Your AI Team (10 agents included free) =

🏗️ **Assistant Trainer** — A meta-agent that trains new agents from a plain job description. Describe what you need and deploy bespoke agents — without writing code.

📡 **AI Radar** — Scans your site's visibility to AI search engines. Checks robots.txt, schema markup, content structure, and technical readiness. Scores 0–100.

✍️ **Content Writer** — Creates, edits, and publishes your posts and pages. Describe what you want, review the draft, and publish.

🔌 **Plugin Assistant** — Creates complete WordPress plugins from natural language descriptions. Generates WPCS-compliant code with Gutenberg blocks.

🔍 **SEO Assistant** — Audits posts for title length, meta description, keyword density, headings, and links. Applies fixes directly.

🔒 **Security Assistant** — Monitors failed logins, flags outdated plugins, audits admin accounts, and checks for modified files.

🩺 **Site Doctor** — Finds database bloat, broken internal links, orphaned content, inactive plugins, and PHP errors. Scores your site A through F.

🧭 **WordPress Assistant** — Your guide to WordPress and Agent Builder. Ask questions, get help with settings, and find the right AI assistant for the job.

🖼️ **Media Assistant** — Generate, edit, and manage images in your media library using AI.

🎨 **Theme Assistant** — Helps you choose and customise WordPress themes. Detects your active theme, reads global styles, and recommends changes.

= WordPress AI Abilities — First-Class Integration =

[Agent Builder](https://agentic-plugin.com/) is one of the first AI plugins to fully integrate with the WordPress AI Abilities API introduced in WordPress 6.9. On older versions, Agent Builder bridge the gap so your AI abilities and tools always work.

== Installation ==

= Minimum Requirements =

* WordPress 6.4 or higher (6.9+ recommended for AI Abilities support)
* PHP 8.1 or higher
* MySQL 8.0 or MariaDB 10.6+

= Automatic installation =

Automatic installation is the easiest option -- WordPress will handle the file transfer, and you won’t need to leave your web browser. To do an automatic install of Agent Builder, log in to your WordPress dashboard, navigate to the Plugins menu, and click “Add New.”

In the search field type “Agent Builder,” then click “Search Plugins.” Once found you can view details about it such as release, rating, and description. Most importantly you can install it by clicking “Install Now”.

= Manual installation =

Manual installation method requires [downloading Agent Builder](https://agentic-plugin.com/download/) and then uploading it to your web server. The WordPress codex contains [instructions on how to do this here](https://wordpress.org/support/article/managing-plugins/#manual-plugin-installation).

= Activation =
1. Find Agent Builder and click **Activate** on the Plugins screen.
2. A **sign-up page** lets you opt in for a free AI API key, or skip to use the Setup Wizard to configure your own provider (OpenAI, Anthropic, Google, Ollama, etc.).
4. If you opt in for a free AI API key, you land directly in the WordPress Assistant chat — ready to go.
5. Navigate to **Agent Builder → Agents** to activate more agents, or go to **Agent Builder → Chat** to start a conversation with any of the free bundled agents.

You can re-run the Setup Wizard at any time from **Agent Builder → Dashboard**

= Updating =

Automatic updates should work smoothly, but we still recommend you back up your site.

= You Choose your own LLM provider =

Works with the AI services you already know:
- Agentic AI (free daily quota and API key)
- OpenAI (ChatGPT)
- Anthropic (Claude)
- Cohere  
- xAI (Grok)
- Google (Gemini)
- Meta Llama
- Mistral
- Ollama (100% local)
- Your provider not in the list? You can add as many LLM providers as you like. 

= Your Data Stays Yours =

- Automated file and database backup and restore
- Full control of read and write permissions
- Supervised mode shows you exactly what agents will change before they do 
- Complete audit log of every action performed by your users and agents

= Want More? =

A plethora of AI agents by the Wordpress community is available at agentic-plugin.com.
Or simpy build your own agents (as many as you need) with the Assistant Trainer.

== Frequently Asked Questions ==

= What is Agent Builder? =

Agent Builder is a native WordPress plugin that lets you create, train and manage powerful AI agents directly inside your WordPress site.

Your agents can write blog posts, optimize SEO, monitor site health, fix broken links, moderate comments, generate reports or anything you ask them to do — all while staying 100% inside your own WordPress environment.

= How is Agent Builder different from other plugins? =

Most AI plugins are just chat interfaces that send every request to a LLM provider.

Agent Builder is different:

- Agents are fully autonomous — they can take real actions (publish posts, edit pages, run WP-CLI, send emails, update plugins on schedule…)
- Everything runs natively inside WordPress (deep integration with core functions, hooks, REST API, user roles, capabilities)
- Your data never leaves your server unless you explicitly connect to an external LLM Provider with an API Key.
- You can create specialized agents for different jobs (Content Writer, SEO Optimizer, Security Guardian, Support Agent…) and orchestrate them together with shared tools.
- Built-in approval and backup workflows with audit logs that keeps everything safe and under your control.

Unlike external frameworks (OpenClaw, LangChain-based tools, etc.) that treat WordPress as just another API, Agent Builder treats WordPress as the primary runtime — faster, more secure, and dramatically simpler to maintain.

= Is Agent Builder free? What about Pro features? =

The entire core plugin and all 10 built-in agents are 100% free. Register for a free Agentic AI key to get 20 chat queries, 5 RAG queries, 1 image generation, and 5,000 TTS characters per day — no credit card required. Free accounts show the last 7 days of conversation history.

Connect your own OpenAI, Anthropic, Google, or other API key for unlimited usage with no daily quotas.

= Where does Agent Builder send my data? =

No data is sent to external services unless you choose a cloud based LLM provider such as Agentic AI, OpenAI, Anthropic or Google etc.

If you register with Agentic AI, your site URL and email are sent to agentic-plugin.com to provision your API key and manage your daily token quotas.

You can also:
- Use local models via Ollama / LM Studio (fully private — nothing leaves your server)
- Bring your own API keys for OpenAI, Anthropic, Google, Mistral, etc.
- All prompts and responses stay between your server and the LLM provider you selected.

= Can AI agents really take actions and is it safe? =

Yes — AI agents can perform real WordPress actions (publish/edit/delete posts, install/update plugins, run WP-CLI commands, send emails, modify files, etc.).

Safety features include:
- Automated file and database table backups.
- Granular capability controls (each agent only gets the permissions you allow)
- Manual approval queues for high and medium risk actions
- Full audit logging of every agent decision and action

We strongly recommend starting agents in supervised mode until you build trust.

= How do I connect my AI Agents to Email, WhatsApp / Slack / Google Sheets / Google PSI? =

[Agent Builder Pro](https://agentic-plugin.com/download/) includes native channel support for WhatsApp, Slack, Google Sheets, Email, and Google PageSpeed Insights.

Extensive [Documentation](https://agentic-plugin.com/documentation/) is available.

= What is the WordPress Abilities API? =

The Abilities API (introduced in WordPress 6.9) is WordPress's native system for plugins to declare what actions they can perform — making those actions discoverable by other plugins, the REST API, MCP clients, and the Command Palette.

Agent Builder integrates in both directions:

1. **Outbound** — All agent tools are registered as WordPress abilities under the `agent-builder/` namespace. Any abilities-aware plugin or client can discover and call them.
2. **Inbound** — Agent Builder queries all registered WordPress abilities from other plugins and makes them available as tools your agents can use. Install an abilities-aware plugin and your agents gain new capabilities automatically.
3. **Extended** — Agent Builder also registers common WordPress operations (get posts, manage media, handle taxonomies) as abilities under the `wp-extended/` namespace, filling gaps that WordPress core doesn't cover yet.

Each registered ability includes risk-level metadata, permission checks, and MCP annotations. On WordPress versions before 6.9, the abilities bridge stays inactive and the plugin works normally.

= Do I need coding skills to use Agent Builder? =

No — the Assistant Trainer lets you create new AI agents by simply describing the job or task the new agent must perform.

Example prompts that work:
- "Create a content writer that publishes 800-word SEO-optimized posts every Monday"
- "Build a security agent that scans for vulnerabilities and emails me daily reports"
- "Make a support agent that answers customer questions and creates support tickets"

Advanced users can still edit agent logic, prompts, tools and capabilities manually.

= Is Agent Builder secure for production sites? =

Yes — we follow WordPress security best practices:
- Backup files and directories
- Capability-based permissions
- Nonce-protected admin actions
- Input sanitization everywhere
- No file operations without explicit approval
- File and row based database backups and restore tools 

Start with a staging site, use approval gates, and monitor audit logs closely.

= Where can I report bugs or request features? =

Use the official support forum on WordPress.org or [report an issue](https://agentic-plugin.com/support/) at https://agentic-plugin.com/

We respond to support tickets within 24–48 hours (faster for confirmed bugs).

== Screenshots ==

1. Setup Wizard – choose your AI provider with pricing comparison and recommendations
2. Setup Wizard – step-by-step API key entry with screenshots for each provider
3. Setup Wizard – connection test and first chat with an AI agent
4. Chat interface – talk to any AI agent with quick actions, markdown support, and file uploads
5. Agents overview – manage all your installed agents in one place - just like plugins
6. Agent configuration – set provider, model, mode, capabilities, and channel permissions per agent
7. Audit log – review every AI action, tool call, and decision with full transparency

== AI Agent Updates ==

Agent Builder can check for newer versions of the AI agents you have installed **opt-in only** — if you explicitly enable it. This expedite security patches and improvements to agent logic, prompts, and tool integrations. Updates mean your AI agents work smarter and benefit from improvements by the Wordpress developer community. 

= What is checked =

On the first visit to **Agent Builder → Agents**, the plugin shows a consent screen explaining what data will be sent and why. You can accept or decline. If you decline, no update checks for newer versions of your AI Agents are ever made.

= When enabled =

Once every 12 hours a request is made to:

`https://agentic-plugin.com/wp-json/agentic/v1/agents/check-updates`

The request body contains:

`{ "agents": [ { "slug": "my-agent", "version": "1.0.0" } ] }`

No personal data, passwords, API keys, user information, or site content are included.

= What is returned =

The server responds with a list of newer versions. Nothing is installed automatically — you review and install each update yourself.

= Resetting your choice =

Deactivating and reactivating the Agent Builder plugin clears your consent choice.

= Privacy =

See our [Privacy Policy](https://agentic-plugin.com/privacy-policy/) for full details.

== External Services ==

This plugin lets you connect to third-party AI services to process chat messages and execute tasks. **No data is transmitted until you choose and configure a provider at the Welcome screen or later in Settings.** If you select Agentic AI (the default provider), you are provided with a free API key. For all other providers, you supply your own API key. The plugin sends conversation messages (user input and system context) to your selected provider's API to receive AI-generated responses.

= OpenAI =
* **Endpoint:** `https://api.openai.com/v1/chat/completions`
* **When used:** When OpenAI is selected as the AI provider in Settings.
* **Data sent:** Chat messages, system prompts, tool definitions, and tool call results.
* **Terms of Service:** [https://openai.com/terms](https://openai.com/terms)
* **Privacy Policy:** [https://openai.com/privacy](https://openai.com/privacy)

= Anthropic =
* **Endpoint:** `https://api.anthropic.com/v1/messages`
* **When used:** When Anthropic is selected as the AI provider in Settings.
* **Data sent:** Chat messages, system prompts, tool definitions, and tool call results.
* **Terms of Service:** [https://www.anthropic.com/terms](https://www.anthropic.com/terms)
* **Privacy Policy:** [https://www.anthropic.com/privacy](https://www.anthropic.com/privacy)

= xAI =
* **Endpoint:** `https://api.x.ai/v1/chat/completions`
* **When used:** When xAI is selected as the AI provider in Settings.
* **Data sent:** Chat messages, system prompts, tool definitions, and tool call results.
* **Terms of Service:** [https://x.ai/legal/terms-of-service](https://x.ai/legal/terms-of-service)
* **Privacy Policy:** [https://x.ai/legal/privacy-policy](https://x.ai/legal/privacy-policy)

= Google (Gemini) =
* **Endpoint:** `https://generativelanguage.googleapis.com/v1beta/models/`
* **When used:** When Google is selected as the AI provider in Settings.
* **Data sent:** Chat messages, system prompts, tool definitions, and tool call results.
* **Terms of Service:** [https://ai.google.dev/terms](https://ai.google.dev/terms)
* **Privacy Policy:** [https://policies.google.com/privacy](https://policies.google.com/privacy)

= Mistral =
* **Endpoint:** `https://api.mistral.ai/v1/chat/completions`
* **When used:** When Mistral is selected as the AI provider in Settings.
* **Data sent:** Chat messages, system prompts, tool definitions, and tool call results.
* **Terms of Service:** [https://mistral.ai/terms/](https://mistral.ai/terms/)
* **Privacy Policy:** [https://mistral.ai/terms/#privacy-policy](https://mistral.ai/terms/#privacy-policy)

= Meta Llama =
* **Endpoint:** `https://api.llama.com/v1/chat/completions`
* **When used:** When Meta Llama is selected as the AI provider in Settings.
* **Data sent:** Chat messages, system prompts, tool definitions, and tool call results.
* **Terms of Service:** [https://llama.meta.com/llama3/license/](https://llama.meta.com/llama3/license/)
* **Privacy Policy:** [https://www.meta.com/privacy/](https://www.meta.com/privacy/)

= Cohere =
* **Endpoint:** `https://api.cohere.com/v2/chat`
* **When used:** When Cohere is selected as the AI provider in Settings.
* **Data sent:** Chat messages, system prompts, tool definitions, and tool call results.
* **Terms of Service:** [https://cohere.com/terms-of-use](https://cohere.com/terms-of-use)
* **Privacy Policy:** [https://cohere.com/privacy](https://cohere.com/privacy)

= Ollama (Local) =
* **Endpoint:** User-configured local URL (default: `http://localhost:11434`)
* **When used:** When Ollama is selected as the AI provider in Settings.
* **Data sent:** All data stays on your local machine. No external network requests are made.

= Agentic Update Checks =
* **Endpoint:** `https://agentic-plugin.com/wp-json/agentic/v1/agents/check-updates`
* **When used:** Only when you the administrator has explicitly opted in update checks on the Agents page. Never called otherwise.
* **Data sent:** Installed agent slugs and version numbers only. No personal data, API keys, or site content.
* **How to disable:** Decline the opt-in prompt, or deactivate/reactivate the plugin to reset the choice.
* **Terms of Service:** [https://agentic-plugin.com/terms-of-service/](https://agentic-plugin.com/terms-of-service/)
* **Privacy Policy:** [https://agentic-plugin.com/privacy-policy/](https://agentic-plugin.com/privacy-policy/)

= Agentic Chat =
* **Endpoint:** `https://chat.agentic-plugin.com`
* **When used:** When Agentic AI is selected as the AI provider (default on new installs with user consent).
* **Data sent:** Chat messages, system prompts, license key (for metering), and site URL. Free API key provided.
* **Daily limits:** Free API key includes 20 chat queries, 50,000 tokens, 5 RAG queries, 1 image, and 5,000 TTS characters per day. Conversation history limited to the last 7 days on the free tier. Upgrade to a paid license or bring your own API key to remove all limits.
* **Terms of Service:** [https://agentic-plugin.com/terms-of-service/](https://agentic-plugin.com/terms-of-service/)
* **Privacy Policy:** [https://agentic-plugin.com/privacy-policy/](https://agentic-plugin.com/privacy-policy/)

= Agentic Knowledge (RAG) =
* **Endpoint:** `https://rag.agentic-plugin.com`
* **When used:** When knowledge files or training data are uploaded via the Knowledge page.
* **Data sent:** Site URL, license key, document content for vector indexing.
* **Terms of Service:** [https://agentic-plugin.com/terms-of-service/](https://agentic-plugin.com/terms-of-service/)
* **Privacy Policy:** [https://agentic-plugin.com/privacy-policy/](https://agentic-plugin.com/privacy-policy/)

= Agentic Image Generation =
* **Endpoint:** `https://imagegen.agentic-plugin.com`
* **When used:** When an agent generates or edits images. Only active if the Media Assistant agent is installed and a user requests image generation.
* **Data sent:** Site URL, license key, text prompts, and optionally a source image (for edit operations). No user-identifying information beyond the license key.
* **Terms of Service:** [https://agentic-plugin.com/terms-of-service/](https://agentic-plugin.com/terms-of-service/)
* **Privacy Policy:** [https://agentic-plugin.com/privacy-policy/](https://agentic-plugin.com/privacy-policy/)

= Agentic Video Generation =
* **Endpoints:** `https://videogen.agentic-plugin.com` (video generation, stitching, trimming, captions, analysis, add audio)
* **When used:** When an agent generates video clips, stitches or edits videos, searches for background music, or adds an audio track.
* **Data sent:** Site URL, license key, text prompts, image data (for image-to-video), video URLs (for editing operations), and audio URLs (for audio mixing).
* **Terms of Service:** [https://agentic-plugin.com/terms-of-service/](https://agentic-plugin.com/terms-of-service/)
* **Privacy Policy:** [https://agentic-plugin.com/privacy-policy/](https://agentic-plugin.com/privacy-policy/)

= AI LLM Costs =
API usage costs are billed directly to you by your selected LLM provider according to their pricing.

== Channels ==
All channel integrations (WhatsApp, Slack, Email, Google Sheets, Google PageSpeed Insights) require **Agent Builder Pro**. Configure at **Agent Builder → Integrations → Channels** after activating your license.

== MCP (Model Context Protocol) ==
MCP server integration requires **Agent Builder Pro** plus the free [Agent Builder MCP companion plugin](https://agentic-plugin.com/downloads/agent-builder-mcp.zip). Once both are active, each AI agent is exposed as its own scoped MCP server, discoverable by Claude Desktop, VS Code, Cursor, and any MCP-compatible client.

Configure at **Agent Builder → Integrations → MCP**.

= MCP Companion Plugin =
* **Download:** https://agentic-plugin.com/downloads/agent-builder-mcp.zip
* **When used:** When MCP Integration is enabled in Pro build and the companion plugin is installed.
* **Data sent:** No data is sent to external services by the MCP integration itself — all communication is between your MCP client and your WordPress REST API using Application Passwords over HTTPS.
* **Privacy Policy:** [https://agentic-plugin.com/privacy-policy/](https://agentic-plugin.com/privacy-policy/)

= WhatsApp =
If enabled the plugin integrates with WhatsApp Cloud API (https://developers.facebook.com/docs/whatsapp/cloud-api) to enable two-way messaging with your AI agents via WhatsApp.

To use this feature:
- You must create a free WhatsApp Business Account and app at developers.facebook.com.
- Provide your own Phone Number ID and Permanent Access Token.
- The plugin sends/receives messages via Meta's servers (no messages stored on your WordPress server unless your agent logic saves them).
- See Meta's WhatsApp Business Terms: https://www.whatsapp.com/legal/business-terms

= Slack =
* **Endpoint:** `https://slack.com/api/*`
* **When used:** When the Slack channel integration is enabled in Settings.
* **Data sent:** Bot token for authentication, outbound messages to Slack channels.
* **Terms of Service:** [https://slack.com/terms-of-service](https://slack.com/terms-of-service)
* **Privacy Policy:** [https://slack.com/privacy-policy](https://slack.com/privacy-policy)

= Email =
* **Endpoint:** Uses WordPress's built-in `wp_mail()` function — no external service is contacted by the plugin directly.
* **When used:** When the Email channel is enabled in Integrations → Channels.
* **Data sent:** Outbound email content is passed to whatever mail transport your WordPress site is configured to use (local sendmail, SMTP plugin, etc.).

= Google Sheets =
* **Endpoints:** `https://oauth2.googleapis.com/token`, `https://sheets.googleapis.com/v4/spreadsheets/*`
* **When used:** When the Google Sheets channel integration is enabled in Integrations → Channels.
* **Data sent:** Service account JWT for OAuth, spreadsheet read/write operations.
* **Terms of Service:** [https://developers.google.com/terms](https://developers.google.com/terms)
* **Privacy Policy:** [https://policies.google.com/privacy](https://policies.google.com/privacy)

= Google PageSpeed Insights =
* **Endpoint:** `https://www.googleapis.com/pagespeedonline/v5/runPagespeed`
* **When used:** When the Google PSI channel is configured in Integrations → Channels and your agents run a website speed audit.
* **Data sent:** The URL being tested and your API key. No personal user data is transmitted.
* **Terms of Service:** [https://developers.google.com/terms](https://developers.google.com/terms)
* **Privacy Policy:** [https://policies.google.com/privacy](https://policies.google.com/privacy)


== Changelog ==

= 2.9.45 - 2026-04-26 =
* Release 2.9.45



= 2.9.45 - 2026-04-26 =
* Release 2.9.45



= 2.9.45 - 2026-04-26 =
* Quick Start page (signup) copy improvements: clearer registration description, corrected consent wording, skip link with provider list on separate line
* Admin menu restricted to Quick Start only when no LLM provider is configured; all other Agentic pages redirect to Quick Start
* Fix: Ollama provider incorrectly reported as configured when only the default URL was present (not explicitly saved)
* Fix: agentic-setup page returned "access denied" when accessed from the Quick Start skip link
* Prerelease script: new checks verify exactly 10 bundled agents are present in library/ and that each agent's README.md version matches agent.php

= 2.9.42 - 2026-04-26 =
* Exception guards with Security_Log logging on all critical execution paths (tool execution, chat, approval execution, job processor tool loop)
* Fix CI release workflow: zip filename mismatch (premium → pro)
* Stop tracking site-specific abilities-signature files in git

= 2.9.38 - 2026-04-25 =
* Cleaner activation: removed silent phone-home on plugin activation — no external requests fire without explicit user consent
* Replaced post-activation hard redirect to sign-up page with a non-intrusive admin notice banner, so CLI-activated installs work correctly
* Sign-up flow now stores a license key returned alongside the API key in a single request, eliminating a separate background HTTP call
* Added directory index.php guard files to all plugin subdirectories to prevent directory listing
* Fix: deactivate → reactivate now correctly shows the setup banner again

For the full version history, visit [agentic-plugin.com/changelog](https://agentic-plugin.com/changelog/).

== Upgrade Notice ==

= 2.9.16 =
Freemium UX improvements: dashboard upgrade nudge, proper free-build integrations page, 7-day history for free tier, clearer quota error messages with upgrade links. MCP and channel code excluded from free build. No breaking changes.

= 2.9.14 =
Internal rename of pro directory and constants. No breaking changes for end users.

= 2.9.13 =
Privacy improvement: agent update checks are now opt-in. No breaking changes.

= 2.9.12 =
Test suite and prerelease tooling fixes. No breaking changes.

= 2.9.11 =
Security hardening (prepared statements), dead code removal, WP.org Plugin Check compliance fixes. No breaking changes.

= 2.9.10 =
WordPress Coding Standards compliance pass, new AI models (Claude 4.x, Gemini 2.5, Grok 3), per-role usage limits. No breaking changes.

= 2.9.9 =
Build pipeline improvements. No breaking changes.

= 2.9.8 =
Free Agentic AI included with daily quota. Privacy improvements. No breaking changes.

= 2.9.5 =
Daily usage quotas, conversations page, 10 built-in agents. No breaking changes.

== Legal ==

By using this plugin you agree to our Terms of Service: https://agentic-plugin.com/terms-of-service/