OpenCollective MCP Server

A Model Context Protocol server that gives AI agents programmatic access to OpenCollective, Hetzner Cloud, and Cloudflare — enabling automated bookkeeping, collective management, and invoice handling without manual intervention.

What is it?

This MCP server exposes 22 tools that let AI agents interact with:

• OpenCollective GraphQL API v2 — Manage collectives, submit expenses, process payments, query transactions, and handle members

• Hetzner Cloud invoices — Automatically fetch, parse, and reconcile hosting invoices

• Cloudflare billing — Track CDN and hosting costs via the billing history API

The Problem

If you run an OpenCollective-backed project with infrastructure on Hetzner, you face a tedious monthly ritual:

1. Log into Hetzner → navigate to invoices → download the PDF

2. Manually extract the amount, date, and line items

3. Log into OpenCollective → create a new expense → copy-paste the data

4. Submit and wait for approval

Repeat every month. Forever.

This MCP automates it.

Who is it for?

• Open collective maintainers who want to automate expense workflows

• DevOps engineers running infrastructure on Hetzner who track costs via OpenCollective

• AI developers building agents that need to manage budgets, expenses, or financial reporting

• Bookkeepers tired of copy-pasting invoice data between systems

What can it do?

OpenCollective Operations (13 tools)

Hetzner Operations (6 tools)

Cloudflare Operations (3 tools)

The Monthly Bookkeeping Workflow

# 1. Fetch the latest Hetzner invoice
invoice = hetzner_get_latest_invoice()

# 2. Submit it as an expense to your collective
oc_create_expense(
    account_slug="my-collective",
    description=f"Hetzner Cloud - {invoice['date']}",
    expense_type="INVOICE",
    payee_slug="my-org",
    items=[{
        "description": f"Cloud services - {invoice['date']}",
        "amount_cents": invoice['amount_cents'],
        "currency": "EUR"
    }],
    tags=["hetzner", "hosting"]
)

# For Cloudflare (automatically converts USD to EUR)
invoice = cloudflare_get_latest_invoice()

oc_create_expense(
    account_slug="my-collective",
    description=f"Cloudflare - {invoice['date']}",
    expense_type="INVOICE",
    payee_slug="my-org",
    items=[{
        "description": f"Cloudflare services - {invoice['date']}",
        "amount_cents": invoice['amount_cents_eur'],  # Already converted!
        "currency": "EUR"
    }],
    tags=["cloudflare", "hosting", "cdn"]
)

That's it. One agent prompt = one booked expense.

What it's NOT

• Not a replacement for human judgment — Expenses still need approval based on your collective's policies

• Not a financial advisory tool — It moves data, not money; you control the payouts

• Not limited to Hetzner — The oc_execute_graphql tool lets you run any OpenCollective operation, so you can manage Stripe payouts, Wise transfers, or budget forecasting

• Not a GUI — It's a backend for AI agents; use the OpenCollective dashboard for manual tasks

Quick Start

1. Clone and install

git clone https://github.com/your-org/opencollective-mcp.git
cd opencollective-mcp
pip install -e .

2. Configure environment

Copy .env.example to .env and fill in:

cp .env.example .env
# Edit .env with your credentials

3. Run the server

# Direct
python -m opencollective_mcp

# Or via installed script
opencollective-mcp

4. Connect to Claude Desktop

Add to your ~/.claude.json:

{
  "mcpServers": {
    "opencollective": {
      "command": "python",
      "args": ["-m", "opencollective_mcp"],
      "env": {
        "OPENCOLLECTIVE_TOKEN": "your-token",
        "HETZNER_ACCOUNT_EMAIL": "you@example.com",
        "HETZNER_ACCOUNT_PASSWORD": "your-password",
        "CLOUDFLARE_API_TOKEN": "your-cloudflare-token"
      }
    }
  }
}

Why build this?

Because infrastructure costs should be visible, automated, and auditable — exactly what OpenCollective provides for open source projects.

We built this to solve our own bookkeeping pain: tracking Hetzner hosting costs for Going Dark and automatically submitting them as expenses each month. Now our AI agent does it.

If you run a collective with cloud infrastructure, this saves you 15–30 minutes every month — and eliminates human error from manual data entry.

License

Apache License 2.0

Built with FastMCP, httpx, and Playwright.