Deterministic pricing your AI agents can trust

Overview

Quotix MCP exposes a company's pricing engine as MCP tools. An AI agent (in CloseBot, n8n, Claude, a custom app…) collects the job details from the customer, then calls Quotix to turn those details into a priced quote.

How it works

1 — Identify the business

Every call carries platform + external_id. Quotix maps that pair to a tenant, then loads the pricing engine for the requested service_type.

(platform, external_id)  →  tenant  →  pricing engine (by service_type)

"gohighlevel" + "ghl_sub_ny_445"  →  Empire HVAC NY  →  "AC Replacement" engine

2 — Discover the required inputs

Call get_quote_schema to get the engine's input_schema (JSON Schema) and its required_fields. Your agent asks the customer for exactly those.

3 — Get the verified quote

Call calculate_quote with the complete inputs. Quotix validates against the schema, runs the deterministic calculation, and returns the total plus a full breakdown.

Connection details

Quickstart

Pick your client. All of them need the endpoint URL and the Authorization header above.

n8n (MCP Client node)

1. Add an MCP Client (or Tool MCP) node.
2. Server URL: https://mcp.quotix.ai/mcp
3. Transport: HTTP Streamable (not legacy SSE)
4. Add header Authorization: Bearer <QUOTIX_API_KEY> (store it as a credential)
5. Select the tool: get_quote_schema or calculate_quote

Claude Desktop / generic MCP client

Clients that speak HTTP transport can point straight at the endpoint:

{
  "mcpServers": {
    "quotix-pricing": {
      "url": "https://mcp.quotix.ai/mcp",
      "headers": {
        "Authorization": "Bearer <QUOTIX_API_KEY>"
      }
    }
  }
}

{
  "mcpServers": {
    "quotix-pricing": {
      "command": "npx",
      "args": ["-y", "mcp-remote", "https://mcp.quotix.ai/mcp",
               "--header", "Authorization: Bearer <QUOTIX_API_KEY>"]
    }
  }
}

Cursor / VS Code

Add the same block to .cursor/mcp.json (project) or your global MCP config.

Python (fastmcp Client)

from fastmcp import Client

async with Client("https://mcp.quotix.ai/mcp", auth="<QUOTIX_API_KEY>") as client:
    schema = await client.call_tool("get_quote_schema", {
        "platform": "gohighlevel",
        "external_id": "ghl_sub_ny_445",
        "service_type": "AC Replacement",
    })
    print(schema.structured_content["required_fields"])

curl (connectivity check)

The MCP endpoint speaks JSON-RPC over HTTP. A quick tools/list:

curl -sS https://mcp.quotix.ai/mcp \
  -H "Authorization: Bearer $QUOTIX_API_KEY" \
  -H "Content-Type: application/json" \
  -H "Accept: application/json, text/event-stream" \
  -d '{"jsonrpc":"2.0","id":1,"method":"tools/list"}'

Tools reference

{
  "engine_id": "b7c1e0a2-...",
  "engine_name": "Residential AC Replacement",
  "service_type": "AC Replacement",
  "tenant_id": "a1b2c3d4-e5f6-7890-abcd-ef1234567890",
  "input_schema": {
    "type": "object",
    "properties": { "tonnage_capacity": { "type": "number" }, "...": {} },
    "required": ["tonnage_capacity", "us_state_location", "site_accessibility"],
    "additionalProperties": false
  },
  "required_fields": ["tonnage_capacity", "us_state_location", "site_accessibility"]
}

{
  "engine_used": "Residential AC Replacement",
  "engine_id": "b7c1e0a2-...",
  "service_type": "AC Replacement",
  "tenant_id": "a1b2c3d4-e5f6-7890-abcd-ef1234567890",
  "total": 6416.25,
  "breakdown": {
    "line_items": [
      { "key": "equipment", "label": "Base Equipment",     "amount": 4900.0 },
      { "key": "labor",     "label": "Installation Labor",  "amount": 1000.0 }
    ],
    "modifiers": [],
    "subtotal": 5900.0,
    "tax": 516.25,
    "tax_rate": 0.0875,
    "total": 6416.25
  }
}

Schema-gate workflow

Use this loop so the agent never invents missing job details.

1. Call get_quote_schema with platform, external_id, service_type.
2. Collect every value in required_fields from the customer.
3. Call calculate_quote with the complete inputs.
4. If it returns VALIDATION_ERROR, read details.missing_fields, ask those questions, and retry.

Errors

MCP tool errors return isError: true with a structured payload. The REST API returns the same body with an HTTP status.

Error payload

{
  "error": "VALIDATION_ERROR",
  "message": "Required quote inputs are missing or invalid.",
  "details": { "missing_fields": ["site_accessibility"] }
}

Also expect standard 401 Unauthorized when the Bearer token is missing or wrong.

Worked example

The Empire HVAC NY pilot tenant. Copy these values to test end to end.

Request — calculate_quote

{
  "platform": "gohighlevel",
  "external_id": "ghl_sub_ny_445",
  "service_type": "AC Replacement",
  "inputs": {
    "tonnage_capacity": 3,
    "us_state_location": "New York",
    "site_accessibility": "Attic - pull-down access"
  }
}

Result

Pricing engines

Each service_type is a JSON engine: an input_schema (what to collect) and a calculation (how to price). Engines are data — new service types are a config change, not a code deploy.

{
  "schema_version": "1.0",
  "display_name": "Residential AC Replacement",
  "shareable": true,
  "input_schema": {
    "type": "object",
    "properties": {
      "tonnage_capacity":   { "type": "number", "minimum": 1, "maximum": 10 },
      "us_state_location":  { "type": "string", "enum": ["New York", "New Jersey", "Connecticut"] },
      "site_accessibility": { "type": "string", "enum": ["Attic - pull-down access", "Rooftop - ladder access", "..."] }
    },
    "required": ["tonnage_capacity", "us_state_location", "site_accessibility"],
    "additionalProperties": false
  },
  "calculation": {
    "line_items": [
      { "key": "equipment", "label": "Base Equipment",
        "formula": "base_equipment + (tonnage_capacity * ton_multiplier)",
        "vars": { "base_equipment": 2500, "ton_multiplier": 800 } },
      { "key": "labor", "label": "Installation Labor",
        "formula": "labor_hours * labor_rate",
        "vars": { "labor_hours": 8, "labor_rate": 125 } }
    ],
    "modifiers": [
      { "when": { "field": "site_accessibility", "equals": "Rooftop - ladder access" },
        "add": 350, "label": "Rooftop access surcharge" }
    ],
    "tax": { "rate": 0.0875 }
  }
}

REST API

The same pricing brain over plain HTTP, for clients that don't speak MCP. Same Bearer key.

curl -sS -X POST https://mcp.quotix.ai/v1/quotes/calculate \
  -H "Authorization: Bearer $QUOTIX_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "platform": "gohighlevel",
    "external_id": "ghl_sub_ny_445",
    "service_type": "AC Replacement",
    "inputs": { "tonnage_capacity": 3, "us_state_location": "New York",
                "site_accessibility": "Attic - pull-down access" }
  }'

Interactive API reference (OpenAPI / Swagger UI): https://mcp.quotix.ai/docs

Request access

Onboarding is done by the Quotix team (no self-serve yet). To get an API key and your tenant configured, send us:

FAQ

Why can the total differ from the Quotix website form?

Expected in V1. The website form builder and this chat-pricing service are separate systems until they're merged. This service is the source of truth for agent-driven quotes.

Can we add a new service type without a developer?

Yes. A service type is a JSON engine (schema + calculation). Add the file and re-seed — no application redeploy.

Is the API key safe to share?

Only with trusted systems (your n8n instance, your team). Treat it like a password. Production and staging use different keys.

Does Quotix generate the PDF / write back to GoHighLevel?

No. Quotix returns price JSON only. PDF proposals and GHL write-back are handled in your n8n workflow (GoHighLevel has its own MCP server for write-back).

Is there an LLM in the pricing path?

No. The agent collects inputs; Quotix validates them against the schema and runs fixed arithmetic. Same inputs → same price, every time.