mcp-recurring
Say "bill Acme 12 hours at 90 EUR on the 1st of every month" once, and stop remembering it. This MCP server stores recurring invoice schedules, client, line items, cadence, start and end dates, and then, when you ask, creates the invoices that have actually fallen due as real records in the invoice server, with its number series, its clients and its A4 PDF. Generation is idempotent: one invoice per schedule per period, keyed by the occurrence date, so running the billing run twice on the same day creates nothing the second time. It also answers "what falls due in the next 30 days" and "what will I invoice per month for the next year". Everything is stored in plain JSON files on your own machine; nothing is uploaded anywhere.
In the official MCP Registry (io.github.theluckystrike/recurring-invoice-scheduler-subscription-billing-due-reminders).
[recurring demo]
Define a repeating invoice once, generate the due PDFs from chat, no billing SaaS required.
60-second install
npm publish for @theluckystrike/mcp-recurring is pending. Until then, the .mcpb one-click bundle or a clone+build
is the working path, both are verified below.
One-click (.mcpb): download recurring.mcpb from the latest release and double-click it in Claude Desktop:
https://github.com/theluckystrike/mcp-servers/releases/latest
(claude_desktop_config.json):
Claude Code:
(.cursor/mcp.json):
The npx form above starts working the moment the package is published. Until then, use the .mcpb bundle above, or
build from source with exactly these three commands:
Then point your client's command at node with one arg: the absolute path to servers/recurring/dist/index.js.
To run in Pro mode set MCP_LICENSE_KEY in the same config block, or call license_activate once with your key.
Pairs with
- mcp-invoice, required in practice, not by code. This server writes into the invoice
server's data directory and shares its number counter, its client list and its business profile, so every
generated invoice appears in
invoice_list, counts in overdue_report and can be re-rendered with
invoice_pdf. Set your issuer details once with business_set there; this server has no business_set
of its own on purpose, so there is only one profile to keep right.
- mcp-time-tracker, for the hours that are not on a retainer. Track them, invoice
them ad hoc, and leave the fixed monthly part to a schedule here.
- mcp-expense-tracker, rebillable costs that change every month belong on an ad
hoc invoice; a schedule is for the amount that does not change.
Tools
Resource: recurring://upcoming returns the next 30 days of occurrences as JSON.
Prompt: monthly_billing_run, dry run, generate, list what is coming, then report who needs a payment reminder.
What you can say
Worked example
The second run is the point: the period, not the calendar day, is the key, so a repeated billing run is a
no-op rather than a duplicate invoice sitting in a client's inbox.
Free vs Pro
Pro is a one-time $19, or $39 for every server in the collection, lifetime.
Dates: what happens at a month end
Every date is a local ISO calendar date, YYYY-MM-DD. An occurrence is the k-th step from start_date, and
occurrence 0 is start_date itself, so a schedule starting today is due today.
weekly = +7 days per step. {days: n} = +n days per step.monthly = +1 month, quarterly = +3 months, yearly = +12 months.- Month ends. The month step keeps the day of month of
start_date and clamps it to the length of the
target month; it never carries the clamp forward. From 2026-01-31 the series is 01-31, 02-28, 03-31,
04-30, 05-31, February does not silently turn a month-end retainer into a 28th-of-the-month retainer.
- Feb 29. The same rule makes a yearly schedule starting
2028-02-29 fall on 02-28 in common years and
back on 02-29 in the next leap year.
anchor_day / end_of_month (Pro). anchor_day replaces the day of month before clamping, so
anchor_day: 31 means the last day of every month; end_of_month: true does the same explicitly. Both are
ignored for weekly and {days: n}, which have no month to anchor to. An anchored first occurrence that
would land before start_date is dropped, never billed early.end_date is inclusive. An occurrence landing exactly on end_date is generated; the next one is not.- Long-lived schedules. Looking up what is due does not replay the schedule from
start_date: it jumps to
an estimate near the date you asked about and scans forward from there, so a daily schedule created in 2010
still reports what is due in 2026 instead of exhausting its per-run occurrence cap walking there one day at
a time.
Money
Amounts are held as integer minor units by the invoice engine, the same ISO 4217 table, the same
round-per-line-then-sum contract, so a schedule's amount and the invoice it produces can never disagree. Each
line's gross is rounded first, tax is computed and rounded per line and grouped into one line per rate, and
the totals are integer sums of those already-rounded values. A schedule bills in its own currency, or your
business default currency if it has none; nothing here converts between currencies.
How it stores data
Schedules and the generation log live in
${XDG_DATA_HOME:-~/.local/share}/mcp-servers/recurring/ as schedules.json and history.json. The
go into the invoice server's directory, ${XDG_DATA_HOME:-~/.local/share}/mcp-servers/invoice/,
with their PDFs under its pdf/ subfolder, the same files invoice_list, overdue_report and
invoice_pdf read there.
Every mutation runs under an advisory lock file. Anything that writes an invoice takes two locks, always
in the same order, recurring/.lock first, then invoice/.lock, so two billing runs (or a billing run
and a hand-written invoice in the other server) cannot interleave, cannot allocate the same invoice number
and cannot deadlock. Invoice numbers are allocated inside the lock; the PDFs are rendered after it is
released, so a slow render never holds up the counter. Saves go to a temporary file and are renamed into
place.
If schedules.json or history.json is unreadable or not valid JSON it is never treated as "empty": the
file is moved aside byte-for-byte as <name>.json.corrupt-<timestamp>, a <name>.json.corrupt marker is
written, and every tool fails loudly until you restore a good copy and delete the marker. This matters more
here than anywhere else in the collection: a history.json silently read as empty would re-bill every period
the schedule has ever covered.
Limits and honest caveats
- Nothing runs in the background. This is a stdio MCP server: it exists while your client runs it. No
daemon, no cron, no email. Invoices are created when you (or the
monthly_billing_run prompt) call
invoice_generate_due. auto_generate is a marker for that prompt, not a scheduler.
- Nothing is sent to the client. The server produces the invoice record and the PDF; delivering it and
chasing payment is still yours to do.
overdue_report in the invoice server tells you who to chase.
- Free tier allows 3 active schedules. Pausing one frees a slot; the paused schedule's history is kept.
- Deleting a schedule keeps its history rows, deliberately: a re-created schedule with the same id cannot
double-bill a period. Invoices already generated are never touched by anything here.
schedule_update changes future periods only. A period already invoiced keeps the amount that was billed;
correct it in the invoice server instead.- No proration and no mid-period cancellation credit: a period is billed in full or not at all.
- No currency conversion; a schedule bills in one currency.
Troubleshooting
npx hangs or fails to find the package: npm publish for this package is pending. Use the .mcpb
bundle or the clone-and-build path above until it lands.- Using the clone path: build
servers/invoice before servers/recurring, the engine is imported from
it. npm run build -w packages/mcp-license -w servers/invoice -w servers/recurring does that in order.
- "No business profile yet": run
business_set in the invoice server (mcp-invoice), not here.
Generation is never blocked by it; the PDF just carries the placeholder issuer "Your business".
- The invoices are not in my invoice server: both servers must see the same
XDG_DATA_HOME. They write
to .../mcp-servers/invoice/ under it; if one client sets that variable and the other does not, you have
two stores.
- A period was skipped:
invoice_generate_due only skips a period already present in history.json.
schedule_history (Pro) or the file itself shows exactly which invoice covered it.
- Node version: requires Node >= 18. Check with
node -v.
Privacy
All data stays local: schedules, the generation log, invoices and PDFs are plain files under your own home
directory. The server makes no network calls at all, and license keys are verified offline.
Built by theluckystrike. MIT. Support: [email protected]
One business profile for the whole suite
Your identity is stored once, at ${XDG_DATA_HOME:-~/.local/share}/mcp-servers/profile/business.json,
and every server in the suite reads it: the invoice issuer, the docx letterhead, the recurring
issuer, expense-tracker's default VAT rate, time-tracker's and timezone's home zone, and the
resume and contract letterheads. Set it once with business_set (invoice or docx) - you never
repeat it anywhere else. An email address is only ever taken from that profile or from an explicit
argument; when none is stored, documents show [add: email] and the tool says so rather than
letting anyone improvise an address.
Use these docs as an MCP server
Any MCP client (Claude, Cursor, Windsurf, VS Code) can read this repository's documentation directly via GitMCP - no install:
