n8n 워크플로를 무제한 Claude로

n8n 워크플로는 일정과 웹훅으로 무인 실행되어 Claude 사용량이 24시간 쌓입니다. 정액 무제한이 빛을 발하는 사용 패턴입니다.

Wire the Anthropic credential

Create an Anthropic credential, paste your unlimited key (sent as x-api-key), and set the Base URL to the gateway host root — no /v1, since n8n appends /v1/messages and /v1/models itself. Attach it to the Anthropic Chat Model node feeding an AI Agent.

A trailing slash on the Base URL breaks the path, and adding /v1 yourself produces a doubled /v1/v1 — host root only.

Fixing 'n8n Claude API key not working'

This common error is usually not a bad key. n8n's credential test calls GET {BaseURL}/v1/models; a gateway without that endpoint fails the test even when chat works. The unlimited gateway implements /v1/models, so the credential saves.

The other cause is auth scheme — n8n sends x-api-key, so a Bearer-only gateway 401s. Use Add Custom Header to send Bearer if required.

Why unattended automation needs unlimited

A single workflow that classifies every inbound ticket, or summarizes a feed hourly, fires Claude calls continuously with nobody watching. Multiply that across a dozen automations and the volume is both large and unpredictable.

On per-token billing a runaway loop at 3am is a real risk. Flat-rate unlimited makes the bill the same whether your workflows make a hundred calls or a hundred thousand — you can schedule everything without fear.

Scaling your automations

Because the cost is fixed, you can add AI steps to workflows you previously kept rule-based, run them more often, and let agents branch into sub-tasks.

One unlimited key covers every Anthropic node across all your workflows for the plan's duration.

# n8n -> Credentials -> Anthropic
# API Key:  <key>
# Base URL: https://claudeapikey.dev   (host root, NO /v1)
# Attach to: Anthropic Chat Model -> AI Agent

FAQ

How do I connect unlimited Claude to n8n?
Create an Anthropic credential with your unlimited key and set the Base URL to the gateway host root (no /v1). Attach it to the Anthropic Chat Model node.

Why is my n8n Claude key 'not working'?
Its credential test calls /v1/models; a gateway lacking it fails the test even though chat works. The unlimited gateway implements it. Bearer-vs-x-api-key mismatch is the other cause.

Why use unlimited for n8n?
Workflows run unattended on schedules and webhooks, generating continuous unpredictable volume. Flat-rate unlimited covers it for a fixed price.

Does the Base URL include /v1?
No — host root only; n8n appends the paths.