The n8n-io/n8n plugin for Claude Code: building and debugging workflows

What n8n-io/n8n's Claude Code plugin actually is

n8n-io/n8n is the plugin Flowy indexes straight from the n8n-io/n8n GitHub repository, the same repo that ships n8n itself. It carries 21 components: 15 skills, 4 agents, and 2 commands, filed under Flowy's automation category. The repository is also one of the most starred projects on GitHub, with roughly 200,000 stars, so this is a large, actively maintained codebase choosing to ship its own Claude Code tooling rather than a small side project.

That 14/4/2 split matters for how you read the rest of this guide. Most of the skills are built for someone assembling or debugging an n8n workflow. Most of the agents are built for someone working on n8n's own codebase.

What is inside

Three skills sit at the center of building something new. workflow-builder handles general workflow assembly: wiring nodes and connections into a working automation. agent-builder is scoped narrower, to n8n's AI Agent node specifically, the part of a workflow that acts on its own rather than following a fixed step order. data-table-manager works with n8n's data table feature, for storing and querying structured data without leaving the workflow.

A second group handles the setup work around a build. setup-mcps configures an MCP server as a tool inside a workflow. credential-recipe-research looks up how a given service expects its credentials configured before you open the credentials screen, and credential-setup-with-computer-use goes further, using Claude's computer use capability to drive that setup screen directly. n8n-cli wraps n8n's own command line interface, and n8n-docs-assistant answers questions against n8n's documentation rather than general knowledge of the product.

The remaining skills cover the lifecycle around a build. intent-recognition works out what kind of request you are making before handing off to the right skill. planning covers scoping the work before it starts, while planned-task-runtime sits closer to n8n's task runner, the external process n8n uses to run Code node scripts outside the main process. post-build-flow covers what happens once a workflow exists, config-evals sets up evaluation of an AI workflow's output against a dataset, and debugging-executions works through a failing or misbehaving execution.

The other four: n8n's own maintainer tooling

The 4 agents point the other way. linear-issue-triager triages issues against n8n's own tracker, and the triage command is the likeliest way to invoke that from a prompt. CHANNEL_INTEGRATION_TESTS and developer read, by name, as internal engineering tooling for the n8n codebase itself rather than for a workflow you are building. prompt-caching reads the same way: about controlling the cost of n8n's own AI-assisted features, not something you configure inside a workflow. The plan command sits closer to the builder side, for scoping a task before starting it.

If you are building on n8n rather than contributing to it, expect to reach for the skills far more often than the agents.

When it pays off

The plugin earns its place when you are actually inside n8n: adding an AI Agent node and wanting agent-builder instead of trial and error, connecting a new service and wanting credential-recipe-research to say what n8n expects before you open the credentials screen, wiring in an MCP server with setup-mcps, or chasing a failed run with debugging-executions. It has little to offer if what you want is general automation advice unrelated to n8n's own node types and concepts.

How do I install it

Flowy indexes this plugin. It does not host, bundle, or maintain it. Install it from the n8n-io/n8n listing on Flowy, which shows the install command for this repository, and every component stays on n8n's own repository and release cadence rather than a Flowy-maintained copy.