pnlcs-mcp
Click on "Deploy Server".
Wait a few minutes for the server to deploy. Once ready, it will show a "Started" state.
In the chat, type
@followed by the MCP server name and your instructions, e.g., "@pnlcs-mcp Which invoices are overdue?"
That's it! The server will respond to your query, and you can continue using it as needed.
Here is a step-by-step guide with screenshots.
pnlcs-mcp
A Model Context Protocol server for PNLCS, the open-source billing panel. Connect Claude Code, Claude Desktop, Cursor, VS Code or any other MCP client to your PNLCS install and work with it in plain English:
"Which invoices are overdue?" — "Show me this client's services and domains." — "Any orders held as fraud today?" — "Open a ticket for ada@example.com about her domain renewal."
Zero dependencies. The server is one small Node process that talks to the PNLCS admin API you already have. Nothing is installed on the PNLCS side.
npm:
pnlcs-mcpRequires: Node 18+ on the machine your AI client runs on
Transport: stdio (spawned by your client — no port, nothing to host)
1. Create an API credential in PNLCS
Log in to the admin area of your PNLCS install.
Go to Configuration → API Credentials.
Click Create, give it a name like
mcp, and copy the two values it shows you:Identifier — the credential's username
Secret — shown once; store it somewhere safe
That pair is all the server needs, passed through three environment variables:
Variable | Value |
| Your install's address, e.g. |
| The credential's identifier |
| The credential's secret |
| Optional. Set to |
Related MCP server: cPanel MCP Server
2. Connect your client
Claude Code (CLI)
One command:
claude mcp add pnlcs \
--env PNLCS_URL=https://billing.example.com \
--env PNLCS_IDENTIFIER=your_identifier \
--env PNLCS_SECRET=your_secret \
-- npx -y pnlcs-mcpAdd --env PNLCS_ALLOW_WRITES=1 if you also want the write tools.
Use --scope project to share the entry with your team via .mcp.json
(put the secret in your shell environment, not in the committed file).
Check it with claude mcp list; remove it with claude mcp remove pnlcs.
Claude Desktop
Edit claude_desktop_config.json
(macOS: ~/Library/Application Support/Claude/, Windows: %APPDATA%\Claude\):
{
"mcpServers": {
"pnlcs": {
"command": "npx",
"args": ["-y", "pnlcs-mcp"],
"env": {
"PNLCS_URL": "https://billing.example.com",
"PNLCS_IDENTIFIER": "your_identifier",
"PNLCS_SECRET": "your_secret"
}
}
}
}Restart Claude Desktop; the tools appear under the hammer icon.
Cursor
Settings → MCP → Add new global MCP server, or create .cursor/mcp.json
in your project (same JSON shape as Claude Desktop above).
VS Code (Copilot agent mode)
Create .vscode/mcp.json:
{
"servers": {
"pnlcs": {
"type": "stdio",
"command": "npx",
"args": ["-y", "pnlcs-mcp"],
"env": {
"PNLCS_URL": "https://billing.example.com",
"PNLCS_IDENTIFIER": "your_identifier",
"PNLCS_SECRET": "your_secret"
}
}
}
}Anything else (Windsurf, Cline, Zed, ...)
Every MCP client that can spawn a stdio server uses the same three pieces:
command npx, args ["-y", "pnlcs-mcp"], and the environment variables
above. From a git checkout, node mcp/server.js works identically.
3. Tools
Read tools — always available
Tool | What it answers |
| Client, order, invoice and revenue totals |
| Health of the install itself |
| Clients, searchable by name/email/company, pageable |
| One client with contacts, by |
| Hosting services of one client |
| Domains of one client |
| Invoices; filter by |
| One invoice with its line items |
| Orders; filter by |
| Support tickets; filter by status |
| One ticket with replies and notes |
| Ticket totals per status |
| Payments, newest first |
| The product catalogue |
| Recent admin and system activity |
Write tools — only with PNLCS_ALLOW_WRITES=1
Tool | What it does |
| Create a client; with |
| Invoice a client with one or more line items |
| Record a payment; marks the invoice paid when covered |
| Open a support ticket |
| Reply to a ticket |
| Suspend a hosting service on its server |
| Lift a suspension |
Without the flag the write tools are not merely hidden — calling one is refused before any HTTP happens. An assistant wired up for reporting cannot even see a suspend button. Give a reporting setup a read-only life by simply not setting the flag.
4. Security notes
The secret only ever travels between the machine running your AI client and your PNLCS install, over the same HTTPS API your admin screens use. It is never sent to the model provider; the model sees tool results.
Prefer a dedicated API credential for MCP so you can revoke it alone.
PNLCS rate-limits API credentials (300 requests/minute per credential), so a runaway agent cannot hammer your install.
Keep
PNLCS_ALLOW_WRITESoff unless you actually want the assistant acting on your books, and read what it proposes before approving tool calls that create or change things.
5. Troubleshooting
Symptom | Cause |
Every tool answers | One of the three variables is missing from the client config |
Every tool answers | Identifier/secret pair is wrong, or the credential was deactivated |
| The install is unreachable from this machine — check the URL and any firewall |
Tools missing in the client | Restart the client after editing its config; check its MCP log for the stderr line above |
Write tools missing | That is the default — set |
6. Tests
npm test runs the offline suite: it spawns the real server process, speaks
the real protocol to it, and checks every tool's HTTP shape against a local
stub — plus the write-gate in both directions, batch requests from older
protocol revisions, version negotiation, timeouts and shutdown draining.
node test/live.mjs runs every tool against a real install — point it at
a demo, never at production books; it creates clearly-named disposable data
for the write tools and prints one PASS/FAIL line per tool.
Development
This package is developed inside the PNLCS monorepo
under mcp/; this repository is its standalone distribution mirror and always
matches the version published on npm.
Available Tools
15 toolsget_activity_logC
Recent admin and system activity.
Inputs: limitnum, limitstart.
get_clientA
One client with contacts, by clientid or email.
Inputs: email, clientid.
get_healthB
Health of the PNLCS install itself.
Inputs: no named parameters.
get_invoiceC
One invoice with its line items.
Inputs: invoiceid (required).
get_statsB
Business overview: client, order, invoice and revenue totals.
Inputs: no named parameters.
get_ticketB
One ticket with its replies and notes.
Inputs: ticketid (required).
get_ticket_countsB
Ticket totals per status.
Inputs: no named parameters.
list_client_domainsC
Domains that belong to one client.
Inputs: clientid (required).
list_clientsA
List clients, newest first. Optional text search and paging.
Inputs: search, status, limitnum, limitstart.
list_client_servicesC
Hosting services that belong to one client.
Inputs: clientid (required).
list_invoicesB
List invoices. Filter by status (draft, unpaid, paid, overdue, cancelled) or client.
Inputs: status, userid, limitnum, limitstart.
list_ordersA
List orders. Filter by status (pending, active, fraud, cancelled) or client.
Inputs: status, userid, limitnum, limitstart.
list_productsC
The product catalogue.
Inputs: no named parameters.
list_ticketsB
List support tickets. Filter by status.
Inputs: status, limitnum, limitstart.
list_transactionsB
Payment transactions, newest first.
Inputs: clientid, limitnum, limitstart.
Tool Schema Changelog
Recent tool additions, removals, and schema changes observed during successful MCP inspections.
Loading tool schema changelog…
TDQS
Scored across 15 tools
Each tool targets a distinct resource or summary view: health/stats/activity are clearly separate from client, invoice, order, ticket, transaction, and product listings. There is little overlap, and the get_/list_ split makes singular detail vs collection retrieval obvious.
Tool names consistently use lowercase snake_case with get_ for single items/details and list_ for collections. The compound list_client_services and list_client_domains still follow the same predictable verb_noun pattern as the rest of the set.
Fifteen tools is at the upper end of the ideal range but every one maps to a distinct and useful view of the PNLCS domain, from clients and invoices to tickets and activity. There is no obvious redundancy or filler.
As a read-only business overview server, the surface is broad and coherent: clients, services, domains, invoices, orders, tickets, transactions, products, stats, health, and activity are all exposed. Minor gaps such as no get_order, get_product, or get_transaction detail views, and no mutation endpoints, keep it from being fully complete.
Maintenance
Related MCP Connectors
Create contacts, move deals, send proposals and invoices, read your inbox, from Claude or ChatGPT.
Manage WordPress blogs and WooCommerce shops from Claude, ChatGPT, Cursor and other MCP apps.
Pennylane MCP server for ChatGPT, Claude and any AI agent: create customer invoice, send invoice, upload supplier invoice, categorise transaction and list unpaid invoices in your own Pennylane account. Cotlet works in the account through your Cloud Browser, a real Chrome where you sign in to Pennylane yourself; Cotlet does not store your password and you need no Pennylane API key. About the official Pennylane API: Yes, Pennylane has a public REST API (Company API v2) with read and write scopes for customers, customer and supplier invoices, transactions, categories and ledger entries (checked 14 Oct 2026). It needs the Essentiel plan or higher and an API token, and the official Pennylane MCP server is read-only for now: it cannot create or send an invoice. Setup and FAQ: https://cotlet.dev/integrations/pennylane-api-mcp-server
Számlázz.hu MCP server for ChatGPT, Claude and any AI agent: issue invoice, cancel invoice, issue receipt, record payment and get invoice PDF in your own Számlázz.hu account. Cotlet works in the account through your Cloud Browser, a real Chrome where you sign in to Számlázz.hu yourself; Cotlet does not store your password and you need no Számlázz.hu API key. About the official Számlázz.hu API: Yes. Számlázz.hu has an official API called Számla Agent, in operation since 2010, that issues, reverses and queries invoices and records payments over HTTP with XML (checked 14 Oct 2026). Setup and FAQ: https://cotlet.dev/integrations/szamlazz-hu-api-mcp-server
Related MCP Servers
- AlicenseNot gradedqualityCmaintenanceEnables AI assistants to manage WHM hosting accounts and server administration tasks including account management, server stats, updates, SSL, backups, and email through a secure API.13ISC
- AlicenseNot gradedqualityDmaintenanceEnables AI assistants to manage cPanel hosting accounts including DNS, email, databases, SSL, files, security, and more through natural language using cPanel's UAPI and API2.ISC
- AlicenseNot gradedqualityCmaintenanceEnables AI assistants to query and manage a web hosting billing platform through its External API — clients, tickets, invoices, domains, orders, products, and system data. It defaults to read-only with per-domain opt-in for writes, two-step confirmation for billing and destructive actions, and automatic redaction of secrets, card data, and PII.1MIT
- AlicenseBqualityBmaintenanceEnables AI agents to autonomously govern cloud hosting infrastructure and billing workflows, including provisioning, invoices, wallet credits, support tickets, and catalog management through 35 tools with JSON:API response hydration.35MIT