AI Langgraph

AI LangGraph connects Drupal’s durable content, users, permissions, and business records to stateful LangGraph workflows through explicit, auditable HTTP contracts.

The project is for teams that want Drupal to remain the control plane and system of record while a separate Python/LangGraph service handles graph execution, model calls, and other long-running integrations. The first contributions will extract reusable patterns from a working reference implementation rather than publish application-specific business logic. APIs may evolve before the first supported release.

Drupal keeps the durable record. LangGraph executes bounded turns. Explicit contracts make each handoff inspectable.

Features

Initial development is organized around a small set of reusable integration capabilities:

• Drupal-first durable state: Drupal owns the business record, permissions, review surfaces, messages, decisions, and integration events that must survive a worker restart.
• Explicit service boundaries: LangGraph runs outside Drupal’s synchronous PHP request lifecycle and communicates through documented, authenticated HTTP contracts rather than a shared application database.
• Server-side credential protection: protected worker requests are proxied by Drupal so private service tokens never need to be exposed to browser JavaScript.
• Replay-safe synchronization: stable operation and idempotency keys make retries observable without silently duplicating messages, state transitions, or external actions.
• Workflow observability: graph traces, durable stopping phases, test journeys, and human feedback can be recorded in Drupal and presented in useful review interfaces.
• Adapter-oriented integrations: model providers, vector stores, queues, calendars, CRMs, and other external systems can remain replaceable rather than becoming hidden dependencies of Drupal content.

Because the project has just been created, this list describes the contribution direction rather than a promise that every capability is already available in a tagged release. Shipped behavior will be accompanied by contracts, tests, and upgrade notes.

Post-Installation

AI LangGraph does not yet have a supported release, so there is no production installation path today. Composer commands, permissions, configuration routes, service health checks, and an end-to-end development example will be documented with the first development release. Until then, follow the issue queue and version-control repository for implementation work.

The intended deployment pattern keeps the LangGraph endpoint and shared credentials in Drupal settings, environment variables, or an appropriate secret manager. Credentials should not be exported with site configuration, stored in content, or sent to the browser. Current development code should be treated as experimental and reviewed carefully before it is used outside an isolated environment.

Additional Requirements

The planned architecture assumes a separately deployed Python service capable of running LangGraph and exposing authenticated HTTP endpoints. Exact Drupal, PHP, Python, and LangGraph version constraints will be pinned in release metadata when the first installable package is published.

A real deployment also needs protected network communication between Drupal and the worker, appropriate secret storage, and durable checkpoint storage when a workflow must continue across multiple turns. LLM providers, vector databases, calendars, queues, and CRM systems are use-case-specific integrations and should remain optional adapters rather than universal requirements.

Recommended modules/libraries

No optional Drupal module is required by a supported AI LangGraph release yet. The AI module is a natural companion when a site needs Drupal-side AI provider abstractions. The AI Agents module is useful when a project also needs Drupal-native, tool-calling agents for content or configuration operations. AI LangGraph is intended to interoperate with this ecosystem, not replace it.

Similar projects

The Drupal AI module provides shared AI APIs, provider integrations, and features for Drupal. AI Agents provides a Drupal-native framework for agents that use tools to inspect or change Drupal configuration and content. AI LangGraph has a different focus: connecting Drupal to state-machine workflows executed by LangGraph outside the PHP request lifecycle while keeping durable business state, access control, audit history, and human review in Drupal.

That boundary is especially useful for multi-turn workflows that pause for human input, retry external services, or resume from a durable phase. The project will document integration points with the wider Drupal AI ecosystem as its APIs stabilize.

Supporting this Module

The most useful support right now is participation. Open issues with concrete use cases, review the proposed contracts, contribute tests and documentation, or share minimal examples of Drupal workflows that would benefit from LangGraph orchestration. Feedback is especially welcome around Drupal access control, queues and retries, idempotency, event schemas, graph observability, and safe multi-turn user experiences.

Before beginning a large implementation, please open an issue so contributors can agree on the reusable boundary first. Contributions should favor framework-level contracts and adapters over assumptions tied to one customer, industry, model vendor, or deployment environment.

most useful support right now is participation. Open issues with concrete use cases, review the proposed contracts, contribute tests and documentation, or share minimal examples of Drupal workflows that would benefit from LangGraph orchestration. Feedback is especially welcome around Drupal access control, queues and retries, idempotency, event schemas, graph observability, and safe multi-turn user experiences.

Before beginning a large implementation, please open an issue so contributors can agree on the reusable boundary first. Contributions should favor framework-level contracts and adapters over assumptions tied to one customer, industry, model vendor, or deployment environment.

Community Documentation

Join the conversation in the Drupal community’s #ai-langgraph Slack channel. The channel grew to 82 members on its launch day, August 24, 2026, and is the place to compare architectures, propose contribution areas, and discuss how LangGraph can complement Drupal’s existing AI ecosystem. Visit Drupal Slack to join the workspace.

For a concrete visualization, explore the FutureStack LangGraph Route Atlas. It maps 32 scheduling test journeys across one LangGraph topology, shows the nodes visited during each turn, and identifies the durable phase handed back to Drupal. The atlas is a reference implementation and design aid—not the release matrix for this contributed module—but it demonstrates the kind of inspectable workflow tooling the community can build together.

Additional architecture notes, setup guides, contract examples, and contributor walkthroughs will be linked here as they are published.