AI & LLM nodes connect n8n workflows with AI products, language-model services, and the systems that support them. Some integrations handle requests to a model. Others connect services for document ingestion, persistent memory, user context, or search. This category helps n8n users find a community node for the step they need. Each listing has its own operations, credentials, and platform requirements, so check those details before building around it.
AI workflows often need more than a prompt and a response. They may need to collect source material, make relevant context available, store useful information for a later run, or keep records organized by person or project. The AI & LLM category includes nodes for those supporting services as well as AI APIs. The current category listing focuses on storing and searching documents and memories, retrieving profile context, and managing related data.
What AI & LLM nodes can do in n8n
Store documents and conversation context
A workflow can send supported text, URLs, files, or conversation data to a connected service for processing. This gives later workflow runs a place to look for information from earlier interactions or approved reference material. The exact accepted inputs and processing steps depend on the node and the service behind it.
For example, a workflow could save a conversation summary after a support interaction, or add a reviewed document to a searchable knowledge collection. A node that supports batch operations can also help move several records in one run. Check the listing for supported formats, size limits, and whether processing finishes immediately or continues in the background.
Find relevant information before a model step
A memory or document search can return context related to a question, topic, or record. Some services support semantic search, keyword matching, or filters that narrow results by metadata or a defined scope. An n8n workflow can pass those results into a separate model step, then route the response for review or delivery.
Search is useful when a model needs information from previous conversations or from a collection of reference documents. It does not make the returned information automatically correct or complete. Review the result fields and test queries against the material you expect the service to find.
Retrieve user or project context
Some AI services keep profile information alongside documents and memories. A workflow can retrieve that context before handling a request, then save a verified update when an interaction adds something useful. This can support personalized assistance or project-aware responses when the service and node expose those operations.
Keep each record associated with a clear user, workspace, or project scope. A consistent tag or container can help separate records during search. Confirm how the connected service applies those scopes, and do not treat a label as an access-control boundary unless its documentation says it is one.
Maintain the information behind an AI workflow
AI-related services may expose actions for reviewing document status, updating records, removing outdated information, or syncing supported external sources. These maintenance steps help keep a knowledge collection useful as the source material changes. The available actions differ by node, and supported external sources are package-specific.
Examples of AI workflows with n8n
The category nodes can sit between the event that starts a workflow and the next step that uses the result. For example:
- Prepare a knowledge collection: receive approved content, add it to a connected service, and check its processing status before making it available for search.
- Answer with relevant context: take an incoming question, search records within the right user or project scope, pass the returned context to a model step, then route the answer through the appropriate review or delivery step.
- Keep useful conversation history: after an interaction, save a concise summary so a later run can retrieve relevant details without relying on the previous execution still being active.
- Refresh stored information: update or remove a record when its source changes, or run a supported sync and route failures for follow-up.
These patterns use other n8n steps around the community node. A category node only performs the actions shown in its listing; it does not automatically provide the trigger, model call, approval, or delivery step around it.
Choose an AI & LLM node that fits the job
Start by naming the part of the workflow you want the node to handle. A direct model request, a memory lookup, document storage, profile retrieval, and source synchronization are different jobs. Look for a listing that exposes the operation you need and returns fields your next workflow step can use.
Before installing, check:
- Supported operations: Can it add, search, retrieve, update, or delete the records your workflow uses?
- Inputs and outputs: Which content types, filters, identifiers, and response fields does it accept or return?
- Authentication: Does it require an API key, OAuth connection, or another credential, and which permissions does that credential need?
- Data scope: Can you separate records by user, workspace, project, or another identifier?
- Processing behavior: Does an action return its result immediately, or should the workflow check a processing status later?
- n8n compatibility: Review the current installation requirements and verification status on the individual listing. Community node requirements vary.
Test with a small, non-sensitive example first. Inspect the returned data and error behavior before connecting the node to a workflow that handles live conversations or important records. Keep credentials in n8n's credential settings, and check what data the service stores before sending it information about users or customers.
Provenance and verification
The current AI & LLM listing is published with npm provenance. Provenance records a link between a package release, its source, and the build that produced it, so you can inspect how that release was created. Provenance and n8n verification are separate. A provenance record does not mean n8n has verified a node. Check the listing for its current verification and installation status; the current listing is marked for self-hosted n8n.
Build an n8n node for your AI product
If you build an AI or LLM SaaS and want n8n users to connect it to their workflows, NativeShip can help you create and publish a community node for your product. Try NativeShip.
