Connect DealScore to n8n

1. In a workflow, add an AI Agent node.
2. Add an MCP Client Tool node and connect its tool output to the AI Agent.
3. Set Server Transport to HTTP Streamable. On node version 1.2 and later that is the default. The current default version is 1.4.
4. Set Endpoint to https://cardealscore.com/api/mcp.
5. Set Authentication to Header Auth and attach the credential below.
6. Leave Tools to Include on All, or switch it to Selected and include score_deal. All also exposes explain_grade and health.
7. Attach your own chat model to the AI Agent. Ask it to call score_deal with a year, make, model, asking price, and mileage. ZIP is optional.

Endpoint: https://cardealscore.com/api/mcp
Server Transport: HTTP Streamable
Authentication: Header Auth
Header name: x-api-key
Header value: YOUR_KEY

Create the key at /keys. A workflow run does not open a browser, so Header Auth is the path to use. Bearer Auth with the same key also works. Full setup: /integrations/n8n.

The MCP Client Tool docs page still names the URL field SSE Endpoint. That is version 1 of the node. Version 1.1 and later call the field Endpoint, and version 1.2 and later add Server Transport. DealScore does not serve SSE, so a version 1 node cannot reach it.

Create a key on /keys. In n8n, create a Header Auth credential. Name is x-api-key. Value is the key. Select that credential on the MCP Client Tool. DealScore needs that one header.

Bearer Auth sends Authorization: Bearer and the same key. DealScore accepts that header on MCP and on POST /api/score. Header Auth is the one that sends x-api-key.

Node version 1.2 and later also lists MCP OAuth2. n8n's MCP OAuth2 credential can register a client with the server (dynamic client registration is on by default) and has an optional Resource URL. Header Auth is the setup on this page.

Why does n8n use an API key instead of OAuth?

A workflow run does not open a browser, so Header Auth is the path to use. The header name is x-api-key and the value is a key from /keys, with no Bearer prefix.

Which MCP Client Tool version do I need?

Version 1.2 or later, where Server Transport can be HTTP Streamable. DealScore does not serve SSE, so a version 1 node cannot reach it.

Does Bearer Auth work in n8n?

Yes. Bearer Auth sends Authorization: Bearer with the same key, and DealScore accepts that header on MCP and on POST /api/score.

Which tools does n8n get, and what do they cost?

score_deal, explain_grade, and health. Free includes 5 scored calls a month. Only results with coverage: scored count, and explain_grade and health are never metered.

Import this workflow, then attach the Header Auth credential and your own chat model. The file has no credentials and no API key.

Download dealscore-n8n-workflow.json

{
  "name": "DealScore score_deal",
  "nodes": [
    {
      "parameters": {
        "content": "## DealScore\n\nThis file has no credentials.\n\n1. Create a Header Auth credential. Name: x-api-key. Value: a key from https://cardealscore.com/keys.\n2. On the DealScore node, choose that credential.\n3. Connect the node to an AI Agent tool input and attach your own chat model.\n\nDo not paste a key into this file.",
        "height": 260,
        "width": 420
      },
      "id": "b7e1c0a2-4d5f-4a61-9c10-0a1b2c3d4e5f",
      "name": "Read me",
      "type": "n8n-nodes-base.stickyNote",
      "typeVersion": 1,
      "position": [
        -40,
        -20
      ]
    },
    {
      "parameters": {
        "endpointUrl": "https://cardealscore.com/api/mcp",
        "serverTransport": "httpStreamable",
        "authentication": "headerAuth",
        "include": "all",
        "options": {}
      },
      "id": "c8f2d1b3-5e60-4b72-8d21-1b2c3d4e5f60",
      "name": "DealScore",
      "type": "@n8n/n8n-nodes-langchain.mcpClientTool",
      "typeVersion": 1.4,
      "position": [
        460,
        80
      ]
    }
  ],
  "connections": {},
  "pinData": {},
  "settings": {
    "executionOrder": "v1"
  },
  "staticData": null,
  "meta": {
    "templateCredsSetupCompleted": true
  },
  "tags": []
}

Example output, illustrative numbers. Score a real listing. A generic 2019 Example Sedan listed at $20,000 with 40,000 miles. The ask is 9% under the expected $22,000 (band $20,000 to $24,000). score_v2 grades that delta_pct Good. The score is 87. comps_count is 20. data_source is example. A live score returns live, mock, or error. These numbers are made up. The same example is on /used-car-price-api. It uses the score_deal response shape. It was not made from n8n, and it does not count against the meter. How the bands are drawn: /guides/how-deal-grades-work.

{
  "year": 2019,
  "make": "Example",
  "model": "Sedan",
  "asking_price": 20000,
  "mileage": 40000
}

{
  "grade": "Good",
  "score": 87,
  "expected_price": 22000,
  "expected_price_low": 20000,
  "expected_price_high": 24000,
  "delta_pct": -0.09,
  "confidence": 0.8,
  "coverage": "scored",
  "rubric_version": "score_v2",
  "data_source": "example",
  "vehicle": {
    "year": 2019,
    "make": "Example",
    "model": "Sedan"
  },
  "comps_count": 20
}

• Server Transport is HTTP Streamable and Endpoint is https://cardealscore.com/api/mcp. SSE Endpoint is the version 1 field, and DealScore has no SSE URL.
• Authentication is Header Auth. Name is x-api-key. Value is the key from /keys, with no Bearer prefix in that value.
• coverage: cannot_score with a null grade means comps were too thin. The connection worked.
• score_quota_exceeded means the monthly cap is spent. The connection worked. Plans: /pricing.

Directory links:

https://cardealscore.com/integrations/n8n?ref=n8n&utm_source=n8n&utm_medium=mcp_directory
https://cardealscore.com/agents?ref=n8n&utm_source=n8n&utm_medium=mcp_directory

Sources: MCP Client Tool, MCP credentials.

score_deal grades a deal. Send asking_price and mileage, plus a VIN or year, make, and model. trim and zip are optional. explain_grade maps a delta_pct to a grade without fetching comps. health reports coverage and limits. Free includes 5 scored calls a month and Builder is $19/mo (founding price) for 1,000 scored calls a month. Only coverage: scored counts, and explain_grade and health are never metered. Plans: /pricing. Argument details: /docs/api. Test arguments in the browser first: /try.

Cursor, VS Code, Claude Code, Claude Desktop, Codex, ChatGPT. The REST call and a recorded response are on /used-car-price-api. Every client is also listed on /agents.