ebill-mcp
Read-only MCP server for the Swiss eBill portal, enabling users to list bills and download invoice PDFs via an authenticated browser session. It cannot approve, release, or pay anything.
README
<div align="center">
ebill-mcp
Read your Swiss eBill archive and download the invoice PDFs. Read-only — it cannot approve, release or pay anything.
</div>
Unofficial, and attached to your bank. eBill is reached only through an authenticated e-banking session, so this server reads through a window you logged into. It is read-only by construction — see What it cannot do — but it is still pointed at a payment system, and it is inherently fragile: the portal is an Angular app that changes without notice. Use it for your own account and respect the provider's terms of service.
An MCP server for the Swiss eBill portal operated by SIX. From any MCP client (Claude Code, Claude Desktop, …) you can list the bills in your archive — biller, amount, due date, delivery date — and download the invoice PDFs.
There is no credential to hold
That is the fact everything else follows from. eBill has no API you can get a key for; the portal is entered by SSO redirect from your bank's e-banking, and the login is second-factor-bound. So a run cannot start without a human.
You log in. The server reads. It opens a visible browser window at your e-banking and hands it to you; it never types into that window. Afterwards it attaches to the same window over the debugging port and reads through the session you established. One login serves many calls — disconnecting a debug client leaves the window standing, which is the whole reason this is usable.
flowchart TD
C["MCP client"] -->|"stdio JSON-RPC"| S["ebill-mcp"]
S -->|"ebill_open"| W["a visible browser window<br/>opens at YOUR e-banking"]
W --> YOU["you log in — second factor,<br/>then open eBill"]
YOU --> TAB[("the eBill tab<br/>stays open")]
S -->|"every other tool"| ATT["attach over the debugging port"]
ATT --> TAB
TAB -->|"fetch with the browser's<br/>own session cookie"| P[("SIX eBill portal")]
ATT --> G{"is the path on<br/>the allow-list?"}
G -->|"no"| NO["refused before it is sent"]
G -->|"yes — five read paths"| P
P -.->|"bills, billers, PDFs"| OUT["answers, and files on disk"]
classDef gate fill:#fff4e5,stroke:#d9822b
classDef bad fill:#fdecea,stroke:#c0392b
classDef you fill:#eafaf1,stroke:#27ae60
class G gate
class NO bad
class YOU,W you
No token ever reaches this process. The session lives in the browser and the browser applies it, so there is nothing here to leak — and nothing to store.
What it cannot do
Approving a bill, releasing one for payment, editing an amount, adding a biller subscription and cancelling a standing approval are all one call away in the same API, and one click away in the same page. "We only wrote reads" is a property of today's code, so it is enforced twice:
- Every request goes through one function that takes no method and no body.
There is no code path in this server that can send anything but a
GET. - Every path is matched against an allow-list first. Five patterns, all reads. Nothing that changes state is on it, so reaching one takes editing the source rather than passing an argument.
Both are tested, and the fixture answers 405 to anything but a GET — so the
assertion "only reads arrived" cannot pass for the wrong reason.
Prerequisites
- Node.js ≥ 20.19
- A Chromium for Playwright, and ideally a real Chrome to log in with
- Swiss e-banking with eBill enabled
git clone https://github.com/sapn95/ebill-mcp.git
cd ebill-mcp
npm install
npx playwright install chromium
Register in Claude Code
claude mcp add ebill --scope user -- node /absolute/path/to/ebill-mcp/index.js
Set your bank's entry page once, so ebill_open does not have to be told every
time — this server does not guess which bank you use:
export EBILL_BANK_URL=https://your-bank.example.com/login
The normal flow
ebill_open— a window opens at your e-banking. You log in and open eBill.ebill_status— confirms the session is reachable and says how long it has left.ebill_list_bills,ebill_download_all, … for as long as the session lasts.- When it expires, every tool says so and names the fix: log in again.
Tools
| Tool | Arguments |
|---|---|
ebill_status |
— |
ebill_open |
url (optional; defaults to EBILL_BANK_URL) |
ebill_settings |
— |
ebill_list_bills |
status (open | done | all), limit |
ebill_get_bill |
bill_id |
ebill_billers |
— |
ebill_download_bill |
bill_id, output_path (absolute) |
ebill_download_all |
output_dir (absolute), status |
Environment variables
| Variable | Default | Purpose |
|---|---|---|
EBILL_BASE_URL |
https://ebill-portal.six-group.com |
portal origin |
EBILL_CDP |
http://localhost:9222 |
where the logged-in window listens |
EBILL_PROFILE |
~/.ebill-mcp/profile |
browser profile — holds a live banking session |
EBILL_BANK_URL |
unset | your e-banking entry page for ebill_open |
EBILL_BROWSER |
chrome |
channel or absolute path |
EBILL_DEBUG |
unset | 1 traces every path read, on stderr |
A variable that is set, even to the empty string, is authoritative: an empty
EBILL_BANK_URL means there is no bank here, not "fall back to the default".
Anything else would make "no bank" impossible to express, and would let a test
that thought it was isolated open a real login page.
Security: the profile directory holds a live e-banking session. It is a secret, it is git-ignored, and it belongs outside any repository. Delete it when a run is done.
Three things the portal does that cost a day to find out
A 200 is not evidence. The Angular server answers any unknown path under
its own prefix with the app shell and a 200 application/json. The four obvious
guesses at a PDF endpoint — payments/{id}/receipt, /document, /bill,
/pdf — all "work" if you read status codes. None of them exists. Every
download here is checked by its first bytes instead.
The PDF is not under /api/. It is
/ebill-portal/ui/payment-pdfs/<businessCaseId>/<receiptFileName>, and the
filename out of the record is part of the path — URL-encoded, not derivable, not
optional. Records without a filename have no document, and are reported as such
rather than given a synthesised name.
One status is half an archive. The bills live under two, done and open,
and the default view shows one of them. ebill_list_bills reads both unless you
narrow it on purpose.
Also worth knowing: businessCaseDate is the date eBill delivered the bill,
which is days to weeks away from the date printed on the document. If you are
filing these, the printed date is usually the one you want.
Checks
npm run gate # lint + smoke + hygiene + tests with coverage enforced
npm test # just the tests
npm run mutate # mutation-test the lines this branch changed
Runs offline and without a bank: the suites attach to a throwaway Chromium holding a local fixture, so nothing they do can reach a real account.
The fixture is the interesting part. It answers the app shell with a 200 for
unknown paths exactly as the portal does — a fixture that returned 404 there
would make the guard against it untestable — and it refuses anything that is not
a GET, so the test asserting that only reads left this process means something.
Mutation testing
npm run mutate asks a different question: not "do the tests pass" but "would
they notice if a guard were removed". StrykerJS deletes one piece of behaviour
at a time and reruns the suite; whatever survives is something no assertion is
watching. It found eleven real gaps in the sibling
pingen-mcp after model review rounds had
stopped turning anything up. Here it runs over the lines a branch changed,
because this suite launches a real browser and a full pass costs hours.
License
MIT © sapn95 — see LICENSE.
Recommended Servers
playwright-mcp
A Model Context Protocol server that enables LLMs to interact with web pages through structured accessibility snapshots without requiring vision models or screenshots.
Magic Component Platform (MCP)
An AI-powered tool that generates modern UI components from natural language descriptions, integrating with popular IDEs to streamline UI development workflow.
Audiense Insights MCP Server
Enables interaction with Audiense Insights accounts via the Model Context Protocol, facilitating the extraction and analysis of marketing insights and audience data including demographics, behavior, and influencer engagement.
VeyraX MCP
Single MCP tool to connect all your favorite tools: Gmail, Calendar and 40 more.
graphlit-mcp-server
The Model Context Protocol (MCP) Server enables integration between MCP clients and the Graphlit service. Ingest anything from Slack to Gmail to podcast feeds, in addition to web crawling, into a Graphlit project - and then retrieve relevant contents from the MCP client.
Kagi MCP Server
An MCP server that integrates Kagi search capabilities with Claude AI, enabling Claude to perform real-time web searches when answering questions that require up-to-date information.
E2B
Using MCP to run code via e2b.
Neon Database
MCP server for interacting with Neon Management API and databases
Exa Search
A Model Context Protocol (MCP) server lets AI assistants like Claude use the Exa AI Search API for web searches. This setup allows AI models to get real-time web information in a safe and controlled way.
Qdrant Server
This repository is an example of how to create a MCP server for Qdrant, a vector search engine.