MCP server

Your agents can run the company. You still hold the pen.

PineMogul speaks the Model Context Protocol. Claude, ChatGPT, Cursor or your own agents use the same commands as the chat and the dashboard, with the same rights and the same yes.

The approval gate

Agents propose. People decide.

A command that changes, sends or spends returns a pending approval instead of acting. The card appears in the feed, on the phone and on the watch.

One command registry for the chat, the dashboard and MCP: nothing has a separate "AI version"
An agent acts with the rights of the person who connected it
Every call is in the audit log with the name of the client
Email and web content is treated as untrusted, to resist prompt injection

# connect from any MCP client
URL   https://mcp.pinemogul.com
Auth  OAuth 2.1 (sign in with your PineMogul account)

Commands

The whole ERP, as typed tools.

Every command has a typed input, a permission and a risk tier. A sample is below; names can still change, and your MCP client shows the live list after you connect.

Jobs

jobs.search
jobs.get
jobs.create_with_crew
jobs.reschedule

Crew

availability.free_members
assignments.set
timecards.list

Assets

assets.search
assets.availability
assets.reserve

Clients

clients.search
clients.get
client_sites.search

Money

invoices.list
invoices.create
invoices.send
payments.list

Documents

documents.search
documents.get

Email

mail.threads
mail.draft
mail.send

Reports

reports.margin
reports.unpaid
audit.query

abc Highlighted commands wait for a yes.

The same care as the chat

Questions are answered at once. Changes need a yes. Money and anything outbound need a yes on an unlocked device.

Live approvals

The agent can wait for the answer, or be told when it arrives.

Your AI allowance is safe

An agent brings its own model. The commands themselves use no AI.

Give your agents a real job.

Start free for 14 days. No card. 30-day money-back guarantee on every plan.

Start free See pricing