Drupal AI Modules by Use Case: What to Install and How Mature Each Is

The AI Dashboard project's index of 8 October 2026 lists 379 contributed modules and 40 recipes in Drupal's AI ecosystem, and 335 of those modules have had a release in the past year. The projects build on the AI (Artificial Intelligence) module, which 24,412 sites report using. They add AI features for content editing, translation and review, image alt text, search answers from site content, AI agents and workflows, and access for external AI assistants.

Drupal founder Dries Buytaert wrote in his State of Drupal post of 7 October 2026: "Drupal AI turns Drupal into an AI harness: organizations can connect AI models to structured content, tools, and workflows, while keeping control through permissions, guardrails, observability, metering, and a choice of AI models."1

This tracker lists one maintained project per use case, 62 in all including providers and vector databases. For each, it gives reported installs, the newest release and security advisory coverage, as checked on Drupal.org on 27 October 2026.

How This Tracker Chooses

• One project per job. Where two projects do the same thing, the tracker lists the one recommended on the AI module's project page; failing that, the one with more active maintenance (recent commits, more maintainers and supporting organisations), then more reported installs, then a stable release covered by security advisories.
• Maintained only. Projects marked deprecated in their code, or obsolete, unsupported or seeking a new maintainer on Drupal.org, are left out, as are projects with no release in the past year and no recent commits.
• Part of the AI stack. Each project uses the AI module or is a companion that the AI module needs or recommends.

The provider and vector database tables list one project per service; a site needs the one for the service it uses.

Reported installs are the counts that sites send through Drupal's Update Status module, so they undercount, favour older projects and are not reported for recipes. Drupal's security advisory policy covers contributed projects that have opted in, and only their stable releases: "That means no security advisories for development releases (-dev), alphas, betas, or release candidates." Eligibility for security advisories is not a security audit, assurance of safety or a guarantee that a vulnerability will be reported. In the tables, "Covered" means an opted-in project with a stable release, "Covered (stable release)" means a newer pre-release on a higher branch is not covered, and "Not opted in" applies whatever the release.

Twenty-eight of the 62 projects are covered, including the AI module, its three most-installed providers, the editing and workflow projects, AI Translate, AI Image Alt Text, AI Logging and AI Metering. Twenty-four have opted in but have no stable release yet: mostly the newer search, agent-tooling and review projects and several provider and vector database connectors, in alpha, beta or release candidate. Separately, ten projects have not opted into coverage at all, five of them with a stable release, including Gemini Provider.

Getting Started on Drupal CMS

On Drupal CMS, the Drupal CMS AI recipe, labelled AI Assistant, installs and configures a set of AI modules: the AI module, AI Agents, AI Dashboard, AI Image Alt Text, a chatbot and Canvas AI. It asks the site builder to choose amazee.ai, OpenAI or Anthropic as the provider; amazee.ai needs no API key for a limited free trial.

The AI Module and Its Dependencies

The AI module provides a common API for calling AI providers. Other modules request a chat response, an embedding, a translation or an image through it, and the AI module sends the request to the provider configured on the site. AI 1.5.1, released on 14 October 2026, ships five active submodules. AI Automators fills fields from other content with 52 automator types, AI Chatbot and the AI Assistant API behind it run chat assistants, AI API Explorer tests calls, and AI Observability records every request. The module's guardrails, configurable checks on AI requests and responses, include input length limits, regular expressions, topic restrictions, a moderation check and a filter for sensitive content in streamed replies.

During 2026 the maintainers moved feature submodules into standalone projects, where, the project page says, they "can evolve faster". AI 1.5.1 marks nine former submodules as deprecated, and this tracker lists their standalone projects instead. One of the nine, AI External Moderation, says in its info file that its functionality was "copied to the AI Core module". Drupal.org change records state that the deprecated AI Search and AI CKEditor copies will be removed in AI 1.6.0. The project page also gives AI 1.3.x an end of life of October 2026.2, 3

The project page sets out the next major version. AI Agents "will be moved into the AI Core module as of 2.0", Tool API "will be a requirement in AI 2.0" and the AI Assistant API "will be deprecated". The AI module's 2.0.x development branch already requires Symfony AI's platform packages. "One major shift in Drupal AI 2.0 is the move towards Symfony AI," AI module maintainer James Abrahams told The DropTimes in a June 2026 interview. "It helps us share the burden of low-level AI code or management of model metadata, so that it is no longer something only Drupal has to manage and becomes part of the PHP community."

Context Control Center, at release candidate since 25 September 2026, stores site-specific context, such as brand voice and editorial rules, and adds it to AI requests. "CCC treats that reusable material as governed Drupal content," its maintainer Kristen Pol told The DropTimes in an August 2026 interview. "At runtime, CCC determines which published context items are relevant to a request."

AI Providers

The AI module connects to "95 and counting" AI platforms, its project page says, through one provider module per vendor or service. The table lists the maintained providers that the AI project page highlights or that report the most installs, including the main cloud vendors, the LiteLLM and OpenRouter gateways, and Ollama for self-hosted models. The three most-installed, OpenAI, Anthropic and amazee.ai, have stable covered releases. The AI module's documentation keeps a matrix of what each provider supports.

Content Editing

These three projects add AI features to the content edit form, and all three are stable and covered. AI CKEditor works inside the text editor, AI Content Suggestions shows suggestions next to the form, and Field Widget Actions adds buttons to individual fields. The buttons can run AI Automators, a submodule of the AI module that fills a field from other content, such as a summary from the body, tags from the text or alt text from an image. Automators run on request, automatically, by cron or in batches.

Translation

AI Translate "adds a button to the translation tab of your content, so you can create a translation with AI in one click", its maintainer Sven Decabooter wrote on his blog on 27 August 2026. Since version 1.4.0, an AI agent asked to translate an article "will use the same code as the button in the interface, including the same permission checks". Sites that manage translation as jobs in the Translation Management Tool (TMGMT) can use AI Translation Management (TMGMT), which adds AI providers as TMGMT translators.

Images, Media and Documents

AI Image Alt Text, with 12,613 reported installs, adds a button that drafts alt text with a vision model for an editor to check before saving, and its bulk submodule updates alt text across existing images. The other three projects have fewer than 160 reported installs each. They suggest focal points for image crops, extract text from office documents and PDFs, and create audio versions of articles.

SEO and Metadata

Metatag AI generates meta tags for the Metatag module, and AI SEO/GEO Analyzer reviews a page for search and generative-engine optimisation and recommends changes. Both are stable and covered.

Search and Answers

"Ask a question and get an AI-generated answer, grounded in the site's own content and with the sources behind it," is how Drupal AI product lead Christoph Breidert described AI search in the Drupal AI Demo announcement of 30 September 2026.

AI Search adds vector embeddings to Search API, so a search can match by meaning rather than exact words, and supplies retrieved content to chatbots and agents. AI powered Search Block, which the AI project page recommends, provides a chat-style search block that answers questions from indexed content. Both are still pre-release: AI Search is in alpha and the block is at release candidate.

Vector Databases

AI Search stores its embeddings in a vector database, through a provider for the database a site runs. The amazee.ai provider in the models table includes one too. The MariaDB and SQLite providers, both stable and covered, store vectors in database software that a Drupal site may already run. Postgres VDB Provider 1.0.0-alpha4, released on 10 September 2026, lists a fix for filter values that were not consistently quoted (issue #3614502) under "Security fix" in its release notes; as an alpha that has not opted in, it gets no Drupal security advisory.

Agents and Page Building

AI Agents runs AI agents that site builders configure in the admin UI. An agent takes instructions and uses tools; the module includes agents that create and change content types, fields and taxonomies. In Drupal Canvas, the Canvas AI submodule adds an assistant with agents that build pages, components and templates and generate titles and metadata. Canvas reports 12,789 installs; the submodule has no count of its own. AI Figma passes the content of a Figma design to Canvas AI's agents.

External AI Assistants and MCP

The Drupal AI Initiative calls the projects above Inside AI. Its Outside AI workstream covers AI agents and assistants that work with Drupal from outside. Dries described the difference when he launched the workstream on 25 June 2026: "with Inside AI, a person uses Drupal, and Drupal uses AI to help. With Outside AI, a person uses an agent, and the agent uses Drupal." Its lead, Scott Falconer, described the aim in an initiative post of 23 July 2026 as "making Drupal legible, callable, safe, and verifiable for agents and builder tools operating from the outside."

The projects in this group build on Tool API, which lets a developer "declare an operation one time — typed inputs, typed outputs, an access check — and then invoke it from Drush, from MCP, from ECA, or from a controller", Tool API maintainer Matt Glaman wrote on his blog on 15 September 2026. Dries called it "one of the most important modules for Drupal developers to watch" in his 7 October post.

MCP Server exposes tools to outside AI clients over the Model Context Protocol (MCP). It is now the protocol runtime only, and its README lists the Tool API bridge, OAuth and admin interface as separate companion projects. The Agent Access recipe installs Simple OAuth, Tool API and MCP Server together, so that people connect AI tools with their own Drupal account and its permissions. Tool Belt supplies ready-made tools for common core actions, Drupal Canvas Tools does the same for Canvas page building, and AI Best Practices gives coding agents Drupal-specific skills. None of these has a stable release yet; Tool API and MCP Server are in beta.

Workflows

ECA runs no-code rules triggered by site events, and AI Integration - ECA adds AI steps and an ECA agent to them. FlowDrop is a visual editor for AI and data workflows, which can pause for input from a person. FlowDrop founder Shibin Das described the split in a July 2026 interview with The DropTimes: "ECA owns the reactive layer and the entry points. FlowDrop owns the advanced, async, long-running, human-in-the-loop layer." All three are stable and covered.

ECA maintainer Jürgen Haas said in an October 2026 interview that AI agents may propose changes to automation but should not approve them: "agents draft, humans deploy."

Review, Validation and Disclosure

In addition to the AI module's guardrails, three projects check content or label AI involvement. AI Validations, stable and covered, adds AI rules to Field Validation, so text or images that break a rule are rejected before save. AI Content Review checks a page against criteria "like brand voice, legal requirements, and reading level", in Breidert's words, and proposes changes for an editor to apply. AI Disclosure records how AI was involved in each item and shows labels for readers, for transparency workflows under Article 50 of the EU AI Act. Both are pre-release.

Logging, Monitoring and Cost

AI Observability, in the AI module, records every AI request, and OpenTelemetry sends those records to monitoring tools. For a simpler log inside Drupal, the AI project page points to AI Logging: "If you feel that AI Observability is too complex to use, this is the module for you." AI Metering, stable and covered, tracks tokens and cost per user and role, with quotas and alerts; when a quota is reached, calls "can gracefully fall back to a local model instead of blocking work", the project page says.

Who Develops the Projects

The Drupal AI Initiative coordinates development of many of these projects. Christoph Breidert leads its Inside AI product work and Scott Falconer its Outside AI workstream, with Marcus Johansson and Artem Dmitriiev as technical leads. On Drupal.org, the AI module lists FreelyGive, Acquia, Soapbox, Velir, Nuvole, QED42, Salsa Digital and 1xINTERNET as supporting organisations.

Projects Not Listed

Eight of the AI module's nine deprecated submodules appear here as their standalone projects, and the ninth, AI External Moderation, now lives in AI core. The Model Context Protocol (mcp) project says it is merging with MCP Server. Orchestration, which connects Drupal to external automation platforms, has had no release since 12 October 2025 and no commits since 22 December 2025.

The AI Dashboard index sorts the wider ecosystem into categories, led by content (71 modules), cloud providers (68), search (54), analytics (51), automation (45), media (44) and chat (37); a module can sit in more than one. For the full lists, the AI project page links the AI Dashboard index and the AI Module Usage listing.

How the Tracker Is Updated

Each table is rebuilt from Drupal.org's release history and project data when the tracker is updated, and the introduction gives the date of the last check. Before putting a pre-release project on a production site, check its project page for the current release and advisory status.4