
Monarch Money
io.github.vargahisPublished by GitHub account vargahisv0.1.4Updated Oct 24, 2026
MCP server for Monarch Money — accounts, transactions, budgets, cashflow, investments, and more
Overview
Lets an assistant read and, if enabled, modify Monarch Money personal finance data such as accounts, transactions, budgets and cashflow.
- What you can do
- Connects an assistant to a Monarch Money account through 44 tools covering accounts, transactions and splits, tags, categories, transaction rules, recurring merchants, budgets, cashflow, investments and credit history. It starts in read-only mode, so write tools are hidden and blocked until explicitly enabled. Read tools retrieve and analyze balances, spending, net worth and budget status; write tools can create, update and delete transactions, tags, categories, rules, budgets and accounts.
- When to use it
- Useful when you want an assistant to answer questions about your own Monarch Money finances, summarize spending or cashflow, or make bulk edits to transactions, tags and rules. Best for personal finance review and analysis rather than general financial advice.
- Requirements
- Runs locally as a Python package via uvx, pip, or a Claude Desktop extension file; desktop only. Authentication happens in a browser on first start, with email, password and a 2FA code if enabled; the resulting token is stored in the OS keyring. No environment variables or headers are declared.
Installation
In SourceWeft
- Open Monarch Money in the dashboard and add it to a workspace.
- Enable the server for the chats that should use its tools.
Desktop only via STDIO. STDIO servers start a local process, so they need the SourceWeft desktop host.
Other MCP clients
Follow the launch instructions in the repository.
README
Monarch Money MCP Server
A Model Context Protocol (MCP) server for integrating with the Monarch Money personal finance platform through Claude Desktop.
Overview
- Secure by design — browser-based login, token stored in OS keychain (never in config files or env vars)
- Safe by default — read-only mode prevents accidental changes; write tools require explicit opt-in
- Comprehensive — 44 tools covering accounts, transactions, splits, budgets, cashflow, tags, categories, transaction rules, recurring merchants, and credit history
- Easy to install — Claude Desktop extension (
.mcpb),uvx, orpip
Two operating modes:
The server starts in read-only mode by default. Write tools are hidden and blocked until you explicitly opt in.
Quick Start
Installation
Option 1: Claude Desktop Extension (.mcpb) — Recommended for Claude Desktop
Enables toggling write mode on/off directly from the Claude Desktop app.
- Download the latest
.mcpbfrom Releases - In Claude Desktop: Settings > Extensions > Advanced Settings > Install Extensions — select the
.mcpbfile - Restart Claude Desktop
To enable write tools: Settings > Extensions > Monarch Money MCP Server > Configure — toggle "Enable write tools" and click Save.
Option 2: uvx (no install required) — Recommended for agents (e.g. Claude Code or Cursor)
Also works with Claude Desktop, but write mode cannot be toggled from the app — set it in the config instead.
Add to your MCP config file:
To enable write tools:
Option 3: pip install — Recommended for local installation and venv
Contributors: See docs/releasing.md for the release process, version scheme, and pre-release testing via TestPyPI.
Add to your MCP config using the full path to your Python interpreter:
To enable write tools, add "--enable-write" to args.
Option 4: Clone and install — Recommended for development
Then add to your MCP config using the Python interpreter from your dev environment:
To enable write tools, add "--enable-write" to args.
Authentication
Authentication happens automatically in your browser the first time the MCP server starts without a saved session.
- Start (or restart) Claude Desktop
- The server detects that no token exists and opens a login page in your browser
- Enter your Monarch Money email and password
- Provide your 2FA code if you have MFA enabled
- Once authenticated, the token is saved to your system keyring — you're all set
Key details:
- Credentials are entered in your browser only — never through Claude Desktop
- Token stored in the OS keyring — persists across restarts, lasts weeks/months
- Expired sessions re-authenticate automatically — the browser login re-triggers on the next tool call
- MFA fully supported
- Fallback: run
python login_setup.pyin a terminal for headless environments
For technical details on the auth architecture, see docs/authentication.md.
Usage Examples
Available Tools
Testing
This project has two complementary test surfaces:
1. Mocked unit tests (the quality gate) — fast, offline, no Monarch connection. These run in CI and must stay green:
The Monarch client is mocked, so these never touch a real account. Live e2e tests (below) are deselected by default.
2. Live end-to-end (e2e) integration tests — exercise the MCP tools against a real Monarch account to verify they handle the live API robustly (adversarial/edge inputs, server-side error paths). They are opt-in and never run in CI:
Prerequisites: a stored keyring token (run python login_setup.py once), or MONARCH_EMAIL /
MONARCH_PASSWORD in the environment. Without these, the suite skips. The tests create and delete
data prefixed with MCP-Test- and self-clean (a post-suite sweep removes any residue). See
tests/integration/README.md for details and safety notes.
There is also a separate agent test skill (
.claude/skills/test-monarch-mcp/) that drives an AI agent to verify it calls the tools correctly — distinct from the two pytest suites above.
🙏 Acknowledgments
Forked from @robcerda's monarch-mcp-server, maintained by vargahis.
Built on the monarchmoneycommunity Python library.
Thanks to:
- @robcerda for the original MCP server
- @hammem for the original monarchmoney library
- @bradleyseanf for the community fork
License
MIT License
Source: README.md at commit 7299fdc
Tools
0Version history
1- v0.1.4LatestSep 16, 2026
Score
50% of the applicable checks met
How complete this listing is, computed from the registry entry, the endpoint and public sources. Not a rating of the server and not a statement that it is safe or working today: reachability is a dated check, and no installs or reviews go into the score.
- Reached in the last 30 days25 pts
Keep the endpoint answering the MCP handshake; SourceWeft probes it every 30 days.
- Auth declared10 pts
Declare the credential headers in the registry entry, or keep the endpoint answering the probe so its auth can be classified.
- Tools listed15 pts
Answer tools/list without a credential, or ship a package that starts in a sandbox without one.
- DescriptionRequired15 pts
- Logo10 pts
- Documentation15 pts
- License10 pts