AI · Model Context Protocol

BillingServ MCP Server
Your billing, one prompt away

Connect Claude, Codex, Cursor, and other AI assistants to your BillingServ account. Ask which customers have unpaid invoices or how revenue is trending, and get answers from live billing data instead of guesses.

What you can ask

Billing questions, answered in plain English

No dashboards, no exports, no digging through reports. Ask your assistant the question you actually have, and it pulls the answer straight from your BillingServ account.

Chase overdue invoices

“Which customers have unpaid invoices this month, and how much is outstanding in total?”

Pull revenue on demand

“Show me the revenue trend for this year and tell me which packages are driving it.”

Spot upgrade opportunities

“What packages does customer 123 have, and what could they upgrade to?”

Raise tickets in context

“Open a billing ticket about the duplicate charge Sarah reported, and add a note to her account.”

Draft invoices and quotes

“Draft a quote for three months of managed hosting and send it to the new customer.”

Check usage before renewal

“How much usage has customer 88 metered this cycle, and is it above their plan?”

Safe by design

Let your AI read freely, keep changes behind approval

Connecting an AI assistant to billing data is only useful if you can trust the boundaries. The BillingServ MCP server is built around a fixed allowlist and a hard split between reading data and changing it.

Reads and writes are separate tools, so you can let your assistant look things up freely while your MCP client asks for approval before anything is created or changed. Some things are not available at all: the server cannot delete records or capture payments, no matter what it is asked.

Fixed endpoint allowlist

The list of BillingServ endpoints the server can call is compiled in and checked on every request. Nothing outside it is reachable.

Reads and writes are separate

Lookups and reports run through one tool, creating records through another. Gate the write tool behind approval in your MCP client and nothing changes without a confirmation step.

No deletes, no payment capture

Destructive endpoints and payment capture are deliberately left out of the allowlist, along with sensitive routes like password reset.

Your API key stays private

The key is read from the environment and only ever sent to the base URL you configure. It is never logged and never included in tool output.

Setup

One config snippet and you’re connected

You need Node.js 20 or newer, your BillingServ API base URL, and an API key. API access is included on every plan. Pick your client, drop in your details, and start asking.

SHELLterminal
# One command, then restart Claude Code
claude mcp add billingserv \
  --env BILLINGSERV_API_BASE_URL="https://billing.example.com/api/v2" \
  --env BILLINGSERV_API_KEY="your_api_key" \
  -- npx -y @billingserv/mcp-server

Using something else? Any MCP client that supports stdio servers works. Point it at npx -y @billingserv/mcp-server with the same two environment variables. Full instructions are in the README on GitHub.

Coverage

What your assistant can see and do

Read access covers the day-to-day questions: customers, invoices, orders, packages, support tickets, usage metering, and the full reporting suite, from annual sales to customer debt.

  • Customers
  • Invoices
  • Orders
  • Packages
  • Reports
  • Support tickets
  • Usage metering
  • Marketing & discounts
  • Tax settings
  • VPN branding & servers

On the write side, your assistant can raise and reply to support tickets, create orders, customers, and notes, draft and send invoices and quotes, send payment reminders, and manage packages and discounts. Every one of those goes through the write tool you can gate behind approval. The full endpoint list is documented in the repository.

FAQ

Frequently asked questions

Answers for teams connecting an AI assistant to their billing for the first time.

What is the BillingServ MCP server?

It is an open-source MCP (Model Context Protocol) server that connects AI assistants like Claude, Cursor, and Gemini to your BillingServ account. Once connected, your assistant can look up customers, invoices, orders, packages, and reports from live billing data instead of guessing, and can raise tickets, orders, and invoices when you ask it to.

Which AI assistants work with it?

Any MCP client that supports stdio servers, including Claude Desktop, Claude Code, Cursor, VS Code with GitHub Copilot, OpenAI Codex CLI, and Gemini CLI. Setup is a single config snippet in each client.

Can the AI change or delete my billing data?

Reads and writes are separate tools, so you can let your assistant look things up freely while requiring approval for anything that creates or changes records. Deleting records and capturing payments are deliberately not available, and the server only calls a fixed allowlist of BillingServ endpoints, checked on every request.

What do I need to set it up?

Node.js 20 or newer, your BillingServ API base URL, and a BillingServ API key. API access is included on all BillingServ plans. Most clients are configured with a single JSON snippet and no separate install step.

How much does the MCP server cost?

Nothing. It is free and open source under the MIT licence, published as @billingserv/mcp-server on npm. You only need a BillingServ account with an API key, and API access is included on every plan.

Is my API key safe?

Your API key is read from the environment and only ever sent to the base URL you configure. It is never logged and never included in tool output. Sensitive routes like password reset are deliberately left out of the allowlist.

Get started

Connect your AI assistant to your billing

Already on BillingServ? Grab your API key and connect your assistant in two minutes. Not yet? Start your 14-day free trial and try it with your own billing data.

Free and open source. MIT licensed.