Skip to main content
GitOps · AI Skills · TypeScript

The AI Skill that gives your coding agent n8n superpowers.

Give your agent the full n8n ontology — nodes, schemas, docs, templates, validation, and the real shape of what can connect to what. Then keep the workflow itself in clean local code, so pull requests stay readable and GitOps stays real.

537 nodes100% schema coverage7,702 templates
workflow lifecycle
$ n8nac list
3 workflows · prod/billing
$ n8nac pull abc123
→ workflows/prod/billing/order-alert.workflow.ts
$ n8nac push order-alert.workflow.ts --verify
schema ok18 nodes · 21 connections
pushedrev 4f1a9c → prod
The case for agentic workflow development

The argument is in the dataset.

Two things at once: a real n8n ontology instead of loose prompts and guesswork, and clean Git-friendly workflow files your team can actually review.

searchable locally in ~5 ms · FlexSearch
n8n nodes indexed
433 core + 104 AI/LangChain. Nothing missing from the official catalog.
537
Schema coverage
10,209 properties and 17,155 option values your agent cannot hallucinate.
100%
Docs pages wired in
93% of nodes have linked official documentation.
1,243
Community templates
The full library, searchable locally in about 5 ms.
7,702
AI nodes
Agents, chains, LLMs, tools, memory, vector stores, retrievers.
104
Example pages
Concrete snippets extracted from official docs for faster repair.
170
Validation before push
Schema checks run before you ship and before your agent drifts.
always
What the stack gives you

A workflow system designed for AI execution and human review.

Every screen below is the real artifact — the terminal output, the file, the diff.

n8nac status
instance prod · project billing
order-alert.workflow.tslocal ahead
weekly-digest.workflow.tsremote ahead
=invoice-sync.workflow.tsin sync

Explicit Git-like workflow

List current status, pull the workflow you want, edit locally, then push a specific filename back to n8n. Clear operations beat hidden background behavior.

skill ▸httpRequest.authentication
predefinedCredentialTypeenum
genericCredentialTypeenum
noneenum
3 of 17,155 option values · no guessing

AI Skill with real context

Schema-accurate node data, documentation, validation, and thousands of workflow examples, packaged for fast local search instead of model guesswork.

order-alert.workflow.ts
@Workflow(name: 'order-alert')
export class OrderAlert {
@Webhook(path: 'orders') trigger;
@If(total > 500) gate;
@Slack(channel: '#alerts') notify;
}

TypeScript workflows

Readable decorators and stable diffs, so humans and coding agents reason about automation as code instead of opaque JSON blobs.

canvas preview
Webhook
If
Slacktrue
NoOpfalse
structure valid · 4 nodes

Canvas preview in VS Code

Edit in code while keeping the n8n canvas close. The IDE becomes the place where review, validation, preview, and sync decisions happen together.

three-way compare1 conflict
base
local
remote
retry: 1
retry: 1
retry: 1
total > 100
total > 500
total > 250
keep localkeep remote

Deterministic conflict resolution

Three-way comparison detects real conflicts, avoids false positives, and keeps resolution visible through diffs and explicit keep-local or keep-remote actions.

instance / project / workflow
workflows/
prod/
billing/
order-alert.workflow.ts
invoice-sync.workflow.ts
staging/
billing/
order-alert.workflow.ts

Project-aware local structure

Instances and projects stay separated on disk, so teams version the right workflows without cross-environment confusion.

What teams actually buy into

Better outputs for agents, better diffs for humans.

01 / agents

For coding agents

Node schemas, option values, docs, templates, and validation instead of guessing how n8n works.

See the agent workflow
02 / gitops

For GitOps

Workflows become readable local files you can diff, review, and merge cleanly in pull requests.

Read the GitOps flow
03 / editor

For VS Code

Inspect status, preview the canvas, validate structure, and push only when you decide to.

Explore the extension
How the loop works

Agentic automation, grounded in GitOps.

The goal is not to hand-author every workflow like application code. It is to let agents build and update workflows from a trustworthy n8n ontology, while your team reviews clean local artifacts instead of opaque UI diffs.

01

Search

Your agent searches nodes, docs, examples, and schemas before it writes anything.

02

Pull

Bring the workflow into a local Git-tracked file so the change is reviewable and reproducible.

03

Edit

Work in JSON or TypeScript with a structure humans read and agents manipulate reliably.

04

Validate

Check against the real schema so bad parameters and fake options get caught early.

05

Push

Ship the exact local file back to n8n with an explicit action, not magical sync.

Start here

Give your coding agent the full n8n map.

Wire an installable n8n ontology into your preferred agent, and keep every workflow change readable, validated, and reviewable from the first pull request.

install
$ npm i -g @n8n-as-code/cli
$ n8nac init
skill installed · ontology ready