automatelab/n8n-nodes-catalog

n8n Nodes Catalog

A structured, machine-readable catalog of n8n node metadata extracted directly from the n8n GitHub repository. Covers 524 nodes across packages/nodes-base (431 nodes) and packages/@n8n/nodes-langchain (93 nodes), sourced from n8n@2.20.6.

Updated monthly. Last updated: 2026-09.

Dataset Summary

This dataset catalogs what each n8n node is: its name, category, supported operations, credential requirements, properties schema, and source location. Existing n8n datasets on HuggingFace (workflow collections, builder training sets) focus on how workflows are assembled. This dataset fills the gap underneath - the node-level metadata that lets an AI agent reason about which nodes to use and what they support, without guessing from stale training data.

Intended Uses

LLM training and fine-tuning. Ground models in current n8n node capabilities. A model that has seen this catalog stops hallucinating node names and operation signatures.

Agent tooling at inference time. An AI agent building an n8n workflow can load this dataset as context to select the right node, check credential requirements, and validate operation names before generating a workflow.

Developer reference. "What n8n nodes support database operations?" is currently a docs-browsing exercise. With this dataset it is a one-liner (see Sample Queries below).

Research. Quantitative analysis of the n8n node ecosystem: coverage by category, credential distribution, operation surface area over time.

Files

Schema

All fields are locked to what the extraction script (extract.py) produces. Do not rely on fields not listed here.

Note on list fields in Parquet: categories, subcategories, group, credentials_required, and operations_supported are stored as JSON strings in the Parquet file (e.g. '["Communication","HITL"]'). Parse with json.loads().

Methodology

The catalog is extracted from the n8n GitHub repository using extract.py. The script:

1. 1.Downloads the n8n release tarball for the target tag (default: latest release). The tarball is cached locally to make re-runs fast.
2. 2.Walks packages/nodes-base/nodes/ and packages/@n8n/nodes-langchain/nodes/, collecting .node.ts files that are NOT inside versioned or implementation sub-directories (v1/, v2/, V1/, V2/, actions/, methods/, transport/, etc.).
3. 3.For each .node.ts, parses the TypeScript source with a targeted regex/AST approach to extract the INodeTypeDescription fields. Also reads the sibling .node.json codex file for category metadata when present.
4. 4.Handles multi-version nodes by reading baseDescription from the primary file and recording defaultVersion. Versioned implementation files (V1/, V2/, etc.) are excluded as they are not standalone nodes.
5. 5.Emits nodes.json (UTF-8 JSON array) and nodes.parquet (Snappy-compressed columnar).

The script is idempotent: re-running with the same tag produces identical output. Run with --tag n8n@2.20.6 to pin to a specific release.

What is not included: credentials definitions, utility modules, the core workflow engine, EE-only nodes that don't follow the standard descriptor pattern.

Update Cadence

This dataset is updated monthly via an automated pipeline. The github_permalink field anchors each record to the specific tag it was extracted from, so older rows remain stable across updates.

The Last Updated field at the top of this card tracks the most recent extraction run.

Sample Queries

Find all nodes that support Slack operations (pandas):

import pandas as pd, json
df = pd.read_parquet("nodes.parquet")
df["ops"] = df["operations_supported"].apply(json.loads)
slack_nodes = df[df["node_name"].str.contains("slack", case=False)]
print(slack_nodes[["display_name", "ops"]])

List all nodes requiring OAuth2 credentials (HuggingFace datasets API):

from datasets import load_dataset
import json
ds = load_dataset("automatelab/n8n-nodes-catalog", split="train")
oauth = [r for r in ds if "oAuth2Api" in json.loads(r["credentials_required"])]
print([r["display_name"] for r in oauth])

Count nodes by category (SQL via DuckDB):

SELECT category, COUNT(*) as node_count
FROM read_parquet('nodes.parquet'),
     UNNEST(json_extract_string(categories, '$[*]')) AS t(category)
GROUP BY category
ORDER BY node_count DESC;

Companion Blog Post

A detailed companion post covering the dataset, queryability, and AI-agent use cases is forthcoming at automatelab.tech. This card will be updated with a direct link when it publishes.

License

Our additions (catalog format, extraction script extract.py, this dataset card, and any editorial framing) are licensed under CC-BY-4.0.

Upstream node metadata: The node metadata in this catalog is derived from n8n source code. Upstream node metadata copyright n8n team, used under the n8n Sustainable Use License. This dataset is a community-maintained catalog/index of that metadata, not a redistribution.

The n8n Sustainable Use License permits derivative works free of charge for non-commercial use and requires preservation of the copyright notice. The full license text is available at https://docs.n8n.io/sustainable-use-license/ and in the n8n repository.

Explicit attribution: Upstream node metadata copyright n8n team, used under n8n SUL. This dataset is a community-maintained catalog/index of that metadata, not a redistribution.

Maintainer

AutomateLab - AI automation guides and tools.