Do these 3 things before closing this tab:
1Clear out junk files and repair common Windows errors2Scan for outdated or missing drivers - takes under a minute3Repair Windows errors before they cause bigger problemsYou can connect Claude to n8n in two ways, and the right one depends on who decides the next step. If Claude should choose which actions to take from a natural-language request, use an AI Agent node with the Anthropic Chat Model node as its model and attach only the tools the agent may use. If the sequence is already fixed and Claude is just one step that returns text, use an HTTP Request node to call the Claude API directly. Both routes need an Anthropic Console API key stored as an n8n credential, and both are billed by token usage on the Anthropic side.
Choose the pattern before you build
The two routes solve different problems. Picking the wrong one is the most common reason an automation ends up either too rigid or too free to act.
As an Amazon Associate I earn from qualifying purchases.
| Question | AI Agent with Anthropic Chat Model | HTTP Request to the Claude API |
|---|---|---|
| Who decides the next action? | Claude chooses from the tools connected to the agent | The workflow order is fixed in advance |
| Typical use | Triage a support email, then decide whether to create a ticket, look up an order, or reply | Summarise every new form submission into a fixed JSON shape |
| Main n8n pieces | Trigger, AI Agent, Anthropic Chat Model, and one or more tool nodes | Trigger, HTTP Request, and the nodes before and after it |
| Main risk | The agent can act through any tool you connect, so tool scope matters most | Less autonomy, but you must handle the request and response format yourself |
| Cost drivers | Every model turn, the tool definitions sent with each request, and tool results returned to the model | The prompt and response sizes of each call you make |
The brief for this guide treats these as implementation patterns inferred from n8n’s documented integration routes. They are not benchmarked comparisons, so validate the choice against your own workflow before relying on it.
Recommended Free Tools
What you need before you start
- An n8n instance you can edit. This can be self-hosted or n8n Cloud. On Cloud, check whether the node you plan to use supports Gateway credits on your plan, because support is not universal.
- An Anthropic Console account with an API key. Claude subscription plans and API access are separate, and API calls are metered.
- A clear list of the actions the workflow may take, written before you open the canvas.
- A low-risk test dataset, such as dummy emails, a sandbox inbox, or a test spreadsheet.
Set up the Anthropic credential
Create the credential once and reuse it across workflows. Keeping the key inside a credential means it does not appear in node parameters, exported workflow JSON, or screenshots you share.
#1 Best Overall
- In the Anthropic Console, create an API key and copy it immediately. Store it in a password manager, since it cannot be read back later.
- In n8n, open Credentials from the left-hand navigation and select Create credential.
- Search for Anthropic, select it, and paste the API key into the API key field.
- Save the credential, then use the connection test if your n8n version offers one. A failed test usually means the key was copied incorrectly or revoked.
Menu labels in n8n change between releases, so if a label differs, follow the Anthropic credential instructions in n8n’s current documentation. The pages checked for this guide were captured between roughly July and late September 2026.
Build an AI agent workflow
The agent pattern works best when the agent has a narrow job and a short list of tools. The steps below assume a manual trigger for testing; swap in your production trigger later.
- Add the trigger. Start with a Manual Trigger so you can run the workflow on demand. Add a Webhook, email, or schedule trigger only after testing is complete.
- Add an AI Agent node and connect the trigger’s output to its main input. Map the incoming text, such as the email body or request field, into the agent’s prompt field.
- Attach the Anthropic Chat Model node to the agent’s model connection. Select your Anthropic credential, then choose a model from the dropdown.
- Attach tools one at a time. Each tool is a separate node connected to the agent’s tool connection. Give every tool a description that says when to use it and what input it expects. The agent reads these descriptions to decide what to call.
- Write a system prompt with boundaries. State the goal, the tools the agent may use, what it must not do, and what to return when it is unsure. For example, tell it to ask for a human review rather than issue a refund.
- Run it manually with low-risk data and open each node’s input and output panel. Confirm that the agent called the tool you expected, with the arguments you expected, before you change anything else.
Tools are the boundary of what the agent can do. If a capability is not connected as a tool, the agent cannot perform it through this workflow, which is useful for keeping scope small.
Free tools Windows power users keep installed
One-click scans. No signup required.
Rank #2
- ALL-IN-ONE HUB & FLEXIBLE MANAGEMENT - Unify your entire smart home setup with a single hub compatible with 2,500+ brands and key protocols including Matter, Zigbee, and Z-Wave. Control all connected devices through a unified dashboard accessible via TV, smartphone, and web apps, plus voice commands using Google Assistant and Alexa.
- PRE-INSTALLED HOME ASSISTANT FOR ADVANCED USERS - Built on Home Assistant, the world’s leading smart home platform, this hub delivers powerful automation, extensive customization, and complete local operation—pre-loaded with HACS and widely supported integrations tailored for experienced users.
- EASY EXPANSION FOR ZIGBEE & Z-WAVE - Simply connect any compatible USB stick—the iSG Box SE automatically recognizes and sets it up. No complex configuration or extra steps required. (Important: This product includes a Zigbee USB stick. Z-Wave USB stick sold separately.)
- BUILT-IN HOMEKIT BRIDGE SUPPORT - Enable Apple HomeKit compatibility for a wide range of smart devices using the integrated bridge. Operate third-party devices that don’t natively support HomeKit directly through the Apple Home app or Siri, without requiring additional accessories.
- LOCAL-FIRST DESIGN & FAST RESPONSE - Experience real-time control thanks to the local-first architecture. Your smart home reacts immediately and stays operational even when the internet is down—free from cloud delays. All data is stored securely within your home network for improved privacy and protection.
Call the Claude API with HTTP Request for fixed steps
When your workflow already knows the order of operations, a plain HTTP Request node is simpler and easier to audit. You are responsible for the request body and for parsing the response.
- Add an HTTP Request node after your trigger.
- Set the method to POST and set the URL to the Messages endpoint shown in Anthropic’s current API reference.
- Add the authentication header using your stored Anthropic credential rather than typing the key into the node. Include the API version header that Anthropic’s reference specifies for the current release, and set the content type to JSON.
- In the body, send the model name, a
max_tokenslimit, and the messages array. Keepmax_tokensas low as the task allows, because output length is billed. - Map the text you need from the response in a following node, such as a Set or Code node, and handle a non-success status with an error branch.
This route does not give Claude any choice of actions. If you later need the model to pick between actions, move the step into an AI Agent node rather than extending a long chain of conditional HTTP calls.
Choose a model without hardcoding a guess
Model names and availability change. Select the model from the dropdown in the Anthropic Chat Model node, or copy the current model identifier from Anthropic’s model documentation at the time you build the workflow. Avoid writing a model name into a long-lived internal note as if it were permanent, and recheck the choice when you upgrade n8n or when Anthropic announces a deprecation.
Understand what the workflow will cost
Anthropic bills tool-use requests by input and output tokens. The input side includes the tools definitions you send with each request, plus the tool-use and tool-result blocks that move between the model and n8n. This means an agent with many tools, long descriptions, or large tool outputs costs more per run than a simple prompt, even when the visible answer is short. Some server-side tools can carry additional usage-based charges.
Do not treat any figure in an article as a current price. Check Anthropic’s pricing page for the model you selected on the day you publish or deploy, and estimate your run cost from a few manual test executions rather than from a general number.
- Cap
max_tokenson every model call. - Connect only the tools the task needs, and keep their descriptions short.
- Return only the fields the agent needs from each tool, not entire API payloads.
- Set a spending limit in the Anthropic Console if your account offers one.
Test before you enable the trigger
Run the workflow manually with dummy data until each step behaves as expected. Look for three things: the agent picks the right tool, the tool receives sensible arguments, and the final output is in the shape the next node expects. Repeat with a few edge cases, such as an empty message, an ambiguous request, or a request that should be refused. Only then switch from the manual trigger to the production trigger.
Run the n8n security audit
An agent that can act is a larger security surface than a workflow that only reads data. n8n provides a security audit that can be run from the CLI, through the API, or from an n8n node. Its checks cover the following areas:
- Credential use, including stale or overly broad credentials
- Risky node types, including SQL query patterns and file-system nodes
- Community and custom nodes you have installed
- Exposed webhooks that accept requests without protection
- Instance settings and whether the instance is up to date
Run the audit before you enable the trigger and again after any significant change. The audit flags areas to inspect; it does not decide whether your workflow is safe. Two practices are editorial recommendations rather than n8n requirements: give each agent only the tools it needs, and add a human approval step before any action that spends money, deletes data, or messages a customer.
Outdated Drivers Are Slowing You Down
One free scan finds every outdated or missing driver and matches the right update for your exact hardware.Free scan · exact hardware matchWindows Errors? Fix Them Before They Spread
Repair common Windows errors and clear accumulated junk for a smoother, more stable PC - no reinstall needed.Free scan · no reinstallProtect any trigger that accepts outside input
A webhook that starts an agent will receive text that may be crafted to manipulate the model. Require authentication on webhook endpoints where n8n supports it, do not pass raw input directly into tools that change data, and keep the system prompt explicit that instructions inside user content do not override the workflow’s rules. Treat this as defence in depth, not as a guarantee.
MCP is a separate option, not a requirement
Anthropic describes the Model Context Protocol (MCP) as an open protocol that standardises how applications provide context to LLMs. The n8n Anthropic Chat Model workflow described above does not require MCP. Use an MCP client or server connection only when your architecture explicitly calls for one, such as sharing the same tools with several AI clients.
Troubleshooting common failures
- Authentication error on the first call. The key is wrong, revoked, or stored in a credential you are not actually using. Create a new key in the Anthropic Console and update the credential.
- The agent answers without calling any tool. The tool description is vague or the request does not match it. Rewrite the description to name the situation in which the tool applies.
- The agent calls the wrong tool. Two tools overlap. Narrow one of them or merge them so the choice is clearer.
- Model not found. The model identifier is outdated. Select a current model from the dropdown in the Anthropic Chat Model node.
- Costs are higher than expected. Check the input and output sizes in the node panels, reduce the number of connected tools, and lower
max_tokens. - The HTTP Request node returns an error. Compare the headers and JSON body with Anthropic’s current API reference, and check the status code and error message in the node output.
Scope of this guide
This guide describes the documented n8n and Anthropic routes as published in mid-to-late 2026. It has not been validated against every n8n version or every Cloud plan. Menu labels, Gateway credit availability, model names, and prices can all change, so verify them in the live n8n editor and on Anthropic’s pricing page before you deploy.
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
Quick wins for a faster PC:
Clear out junk files and repair common Windows errorsFree Scan →Fix the driver behind crashes, sound loss and screen glitchesFind Drivers →
