Locus product guide · Updated
n8n MCP: Client Tool setup, server access and costs
n8n works with MCP in both directions. To give an n8n AI Agent paid tools, attach an MCP Client Tool node, set Endpoint to https://api.paywithlocus.com/api/credits/mcp, Server Transport to HTTP Streamable and Bearer Auth to a Locus Pro agent key. An Apollo people search then costs 2 credits, about $0.002. To let outside AI clients run your workflows, use n8n's instance-level MCP server instead.
Setup, step by step
Step 1
Create an agent key for the workflow
In the Locus Pro dashboard, open Connect, then API keys, then An agent. Pick the tools the workflow may call, a credit cap per call and an expiry. Give each workflow its own key so you can revoke one without breaking the rest.
Step 2
Save the key as an n8n Bearer Auth credential
In n8n, create a Bearer Auth credential and paste the lcac_ key into Bearer Token. n8n sends it as the header Authorization: Bearer followed by the key.
Step 3
Attach an MCP Client Tool to your AI Agent
Add an AI Agent node with a chat model, then connect an MCP Client Tool node to its Tools input. Set Endpoint to the URL below, Server Transport to HTTP Streamable and Authentication to Bearer Auth with the credential from step 2.
https://api.paywithlocus.com/api/credits/mcp
Step 4
Expose only the Locus tools you need
Set Tools to Include to Selected and pick search_apis, describe_api, execute and get_balance. Under Options, raise Timeout above its 60,000 ms default if a slow provider call times out.
Step 5
Test from the chat panel
Send a message that names an endpoint and a ceiling. The execute result reports credits_charged and credits_balance.
Use apollo/people-search to list 5 people at apple.com. Pass max_credits 5 and tell me the credits charged.
Step 6
For a fixed paid step, use the MCP Client node
The MCP Client node runs one MCP tool as a regular workflow step, with no model deciding. Choose the execute tool and Input Mode JSON, then paste arguments like these. The ceiling lives in the workflow, so nothing can raise it at run time.
{"slug": "apollo/people-search", "args": {"q_organization_domains_list": ["apple.com"], "per_page": 5, "page": 1}, "max_credits": 5}
What does n8n MCP mean: client or server?
n8n’s MCP features split by direction. Two nodes make n8n a client that calls outside MCP servers, such as Locus Pro. Instance-level access and the MCP Server Trigger make n8n a server that outside AI clients call. Pick by which side starts the call.
| Feature | Direction | What it does | Where you set it up |
|---|---|---|---|
| MCP Client Tool node | n8n calls an outside server | Gives an AI Agent node the tools of a server such as Locus Pro | Attached to the AI Agent Tools input |
| MCP Client node | n8n calls an outside server | Runs one MCP tool as a normal step | Anywhere in a workflow |
| Instance-level MCP access | Outside clients call n8n | Lets Claude, ChatGPT and others search, run and build enabled workflows | Settings, Instance-level MCP; the URL ends in /mcp-server/http |
| MCP Server Trigger node | Outside clients call one workflow | Exposes the tools of a single workflow | Inside that workflow |
Why does my MCP Client Tool only show an SSE Endpoint field?
That node is version 1, which speaks SSE only. In n8n's source code, version 1 shows SSE Endpoint, and version 1.2 and later show Endpoint plus a Server Transport option that defaults to HTTP Streamable. The docs page for the node still lists only SSE Endpoint.
Locus Pro uses Streamable HTTP, so a version 1 node cannot connect to it. Update n8n and add a fresh MCP Client Tool node, or use the MCP Client node, whose docs list Server Transport.
Bearer Auth or MCP OAuth2 for Locus Pro?
Use Bearer Auth with an agent key. Workflows run on schedules and webhooks with nobody at a browser, and Locus Pro gives unattended code a scoped key with its own tool list, per-call cap and expiry.
The node also offers MCP OAuth2 and Multiple Headers Auth. Locus Pro supports browser OAuth for interactive clients, but we have not tested n8n's MCP OAuth2 credential against it, so this guide does not recommend it.
What does an n8n workflow with paid tools cost per month?
You pay n8n for workflow executions and Locus Pro for tool calls. n8n prices by executions, regardless of how many steps or tool calls run inside one. Locus Pro charges each call its catalog price, and 1,000 credits equal $1. Prices checked November 10, 2026.
A daily lead workflow runs 30 times a month. Each run makes 1 Apollo people search (2 credits) and 10 Hunter company lookups (101 credits). That is 103 credits a run and 3,090 credits a month, about $3.09 in Locus credits, using 30 of the 2,500 executions in n8n Starter.
| n8n plan | Price, billed annually | Workflow executions |
|---|---|---|
| Starter (cloud) | 20€ a month | 2,500 a month |
| Pro (cloud) | 50€ a month | 10,000 a month |
| Business | 667€ a month, self-hosted only | 40,000 a month |
| Community Edition | Self-hosted from GitHub | Not listed on the pricing page |
How do I stop an n8n agent from overspending?
Set the per-call cap on the agent key. Locus enforces it before the provider is called, and a call over it fails with agent_connection_call_limit_exceeded. Limit the key’s tool list to the endpoints the workflow needs.
In n8n, Tools to Include keeps the agent to the four Locus tools. For steps where the model should not choose the price, use the MCP Client node with max_credits fixed in its JSON. Each Locus Pro plan also limits paid calls a minute: 5 on Free, 15 on Plus, 240 on Enterprise and 1,200 on Enterprise Plus.
Which errors come up when n8n calls Locus Pro?
Locus refusals come back as tool results the agent can read. Connection problems fail the node.
| Symptom | Likely cause | Fix |
|---|---|---|
| Node cannot connect | Version 1 node or SSE transport selected | Use a current node with HTTP Streamable |
| 401 Unauthorized | Wrong credential or expired key | Check the Bearer Auth credential |
| credential_endpoint_denied | Endpoint not on the key’s tool list | Add the endpoint to the agent key |
| charge_ceiling_exceeded | Price above max_credits | Raise the ceiling; nothing was charged |
| Tool call times out | Slow provider and the 60,000 ms default | Raise the Timeout option |
Should an n8n workflow use Locus Pro or each provider’s own node?
n8n already has nodes and credentials for many apps. Keep using them for apps your team owns accounts with, such as Slack or a CRM.
Use Locus Pro for paid data the workflow needs only now and then, or from several providers: enrichment, search, scraping or market data. You get one key, one balance and a ceiling on each call, with no extra provider accounts. At high steady volume on one provider, that provider can cost less per call with your own account and API key.
Frequently asked questions
Can Claude or ChatGPT run my n8n workflows over MCP?
Yes, through instance-level MCP access. An owner or admin turns it on in Settings, Instance-level MCP, and each workflow must be enabled for MCP. Clients connect with OAuth or an API key.
Do MCP tool calls use up extra n8n executions?
No. n8n prices by workflow executions with unlimited steps, so tool calls inside one execution do not count as more executions. Locus Pro bills each paid call separately in credits.
Does Locus Pro replace my n8n credentials for every app?
No. It covers the APIs in the Locus catalog. Apps like Slack, Gmail or your CRM still use their own n8n credentials and nodes.
Which n8n version do I need for HTTP Streamable?
Any version that offers MCP Client Tool node version 1.2 or later. If the node only shows an SSE Endpoint field, update n8n and add a fresh node.
Implementation references
Use these first-party references for current request contracts and account requirements. Tool availability, prices, and negotiated terms can change.
Sources
Third-party details were checked on November 10, 2026. Vendors change pricing and features often, so confirm on their site before you decide.