NewModel Context Protocol

Connect approved AI clients to your operational Spine.

The North Ark MCP Server lets Claude, Cursor, custom agents, and other MCP-compatible clients work with live context across your leads, jobs, invoices, tasks, milestones, and client communications.

Get early access See Spine AI instead

What it is

North Ark becomes the service business context your AI connects into.

MCP (Model Context Protocol) is an open standard that lets AI clients read context from and take actions inside external systems. The North Ark MCP Server exposes your Spine — customers, leads, jobs, tasks, quotes, invoices, and communication history — as live operational context.

This means the AI you already use can answer operational questions like what needs action today, which lead has gone quiet, which job milestone is blocking billing, or why an invoice is being disputed.

Every action the AI takes through MCP goes through the same approval and audit infrastructure as Spine AI. One system. Multiple AI surfaces.

“What jobs are overdue right now?”

Calls list_open_jobs with status filter — returns live data from your Spine.

“Create a job for the Henderson site visit on Friday.”

Calls create_job with the details — job appears in your Spine immediately.

“Draft a chase for every invoice over 30 days.”

Calls get_outstanding_invoices, then draft_communication for each — queued for your approval.

“Give me a summary of this week's activity.”

Calls summarise_activity — returns a plain-English business briefing.

Server specification

Resources, tools, and prompts.

The North Ark MCP Server exposes your operating environment through three MCP primitives — resources for live context, tools for taking action, and prompts for common operational workflows.

Resources — live business context

customers

Full customer records, history, and linked jobs

leads

Open leads and opportunities with status and context

jobs

All jobs — open, in progress, completed — with assignees and workflow state

tasks

Tasks with due dates, assignments, and completion status

quotes

Quotes with line items, status, and linked customer

invoices

Invoices, payment status, and outstanding balances

Tools — actions the AI can take

list_open_jobs

Return all open jobs, optionally filtered by assignee, customer, or status

create_job

Create a new job record with customer, description, date, and assignee

update_job_status

Move a job to a new workflow state

create_customer

Create a new customer record with contact details

list_overdue_tasks

Return all tasks past their due date

create_task

Create a task attached to a job or customer

get_outstanding_invoices

Return all unpaid invoices, optionally filtered by age

draft_communication

Draft an email or message for a customer record — queued for approval

list_open_leads

Return all active leads with last activity and status

convert_lead_to_job

Convert a qualified lead into a job record

assign_team_member

Assign or reassign a team member to a job or task

summarise_activity

Return a plain-English summary of business activity for a time period

Prompts — pre-built operational workflows

daily_briefing

What needs doing today — open jobs, overdue tasks, and leads that have gone quiet

invoice_chase

Draft follow-up messages for all unpaid invoices over a specified number of days

job_status_report

Summary of every open job with current state and next action

lead_pipeline

Overview of the current lead pipeline with value estimates and next steps

North Ark is the environment AI connects into — not a tool beside it.

Most AI tools sit beside your operating system. You open them, ask a question, and then manually take the output back into your CRM, your inbox, or your spreadsheet.

The MCP Server reverses that. Your AI client connects directly into North Ark. It reads your real business data, takes real actions, and operates with full context — from whatever AI surface you prefer.

Use Claude Desktop for your daily briefing. Use a custom agent for invoice chasing. Use Cursor to build workflows that interact with your business data. All of it runs through the same North Ark environment, the same approval layer, and the same audit log.

Any MCP-compatible AI client

Claude, Cursor, custom agents, or any tool built on the MCP standard.

Live business data

Every query returns current data from your Spine — not a stale export or a cached snapshot.

Same approval layer as Spine AI

Communications, quotes, and critical changes go through the same queue regardless of which surface initiated them.

Full audit log

Every MCP action is recorded — what was called, what changed, and who approved it.

No lock-in to one AI

Use whichever AI client works best for your workflow. North Ark provides the operating layer underneath.

Getting started

Connect in three steps.

Start with the Spine

Your North Ark environment is the data source. The MCP Server requires an active Spine with your business data loaded.

Generate your MCP key

From your North Ark settings, generate an MCP API key with the scopes your agent needs — read-only, read-write, or specific record types.

Connect your AI client

Add the North Ark MCP Server URL and your key to your AI client configuration. Claude, Cursor, or your own agent can then read and act on your business data.

claude_desktop_config.json

{
  "mcpServers": {
    "north-ark": {
      "url": "https://mcp.northark.ai",
      "apiKey": "your-mcp-api-key",
      "scopes": ["jobs:read", "jobs:write",
                 "customers:read", "tasks:write",
                 "communications:draft"]
    }
  }
}

Make your operational Spine accessible to approved AI clients.

The MCP Server is available to early-access businesses that want AI clients to work with approved service business context and actions.

Get early access See the Spine →