Overview

This MCP (Model Context Protocol) server provides a bridge between LLM-based agents and the Monzo banking API. It exposes a set of tools that let a model or developer perform common Monzo workflows—check balances across personal, joint and flex accounts, list and annotate transactions, and manage pots—using natural language or explicit tool calls. The server implements an MCP-compatible interface so LLMs can discover and invoke these tools during multi-step reasoning.

For developers building chat assistants, automations, or experimentations with agentic workflows, the Monzo MCP server removes the friction of wiring up OAuth, formatting API calls, and handling Monzo-specific data models. It centralizes account access, provides a small, focused toolset, and returns structured results an LLM can consume or a developer can use directly.

Features

Exposes MCP-compatible tools for interacting with Monzo accounts
Read balances across personal, joint, and flex accounts
List, filter, and annotate transactions (add contextual notes)
Create and manage pots; transfer money to/from pots
OAuth integration with Monzo (token refresh and secure configuration)
Returns structured JSON suitable for LLM consumption or programmatic use
Lightweight Node.js service with a simple HTTP interface

Installation / Configuration

Clone the repository and install dependencies:

git clone https://github.com/BfdCampos/monzo-mcp-bfdcampos.git
cd monzo-mcp-bfdcampos
npm install

Create a .env file in the project root with your Monzo API credentials and server settings. Typical environment variables:

# Monzo API credentials
MONZO_CLIENT_ID=your_monzo_client_id
MONZO_CLIENT_SECRET=your_monzo_client_secret
MONZO_REDIRECT_URI=https://yourapp/callback

# A long-lived refresh or access token (if using a personal token)
MONZO_REFRESH_TOKEN=your_monzo_refresh_token_or_token

# Server / MCP settings
PORT=3333
MCP_API_KEY=optinal_api_key_for_local_auth

Start the server (development):

npm run dev
# or for production
npm start

By default the server listens on the configured PORT (3333 if unset). Secure the process and the .env file when running in production.

Available Tools

The server exposes a set of tools (functions) an MCP-compatible LLM or client can invoke. Each tool returns structured JSON and, where relevant, human-readable summaries.

Tool name	Purpose	Key parameters
check_balance	Get balances for an account or all linked accounts	account_id / account_type (personal, joint, flex)
list_accounts	List available Monzo accounts with metadata	none
list_pots	List pots for an account	account_id
create_pot	Create a new pot	name, target_amount (optional)
transfer_to_pot	Move funds into a pot	amount, pot_id, account_id
transfer_from_pot	Move funds from a pot back to account	amount, pot_id, account_id
list_transactions	Fetch transactions with filters (date range, merchant, min/max amount)	start_date, end_date, merchant
annotate_transaction	Attach a note or tag to a transaction	transaction_id, note, tags

Example: using the check_balance tool returns a JSON object with account identifiers, current balance, currency and a brief summary string for display.

Available Resources

GitHub repository: https://github.com/BfdCampos/monzo-mcp-bfdcampos
Monzo Developer Docs: https://docs.monzo.com/ (for API, scopes, and OAuth)
MCP (Model Context Protocol) specification (refer to your MCP client library docs) — the server follows a simple MCP tooling pattern so LLMs can call named tools and receive structured outputs.

Use Cases

Quick balance check (natural language)
- Prompt: “What’s my total balance across personal and joint accounts?”
- Flow: LLM invokes check_balance for each account; server returns per-account balances and a consolidated total.
Move money into a savings pot
- Prompt: “Move £100 from my personal account into the ‘Holiday’ pot.”
- Flow: LLM calls list_pots to find the pot ID, validates funds with check_balance, then calls transfer_to_pot. Server returns transaction ids and status.
Investigate and annotate a transaction
- Prompt: “List transactions at coffee shops in March and label the latte purchase as ‘business expense’.”
- Flow: list_transactions with merchant filter, LLM selects the correct transaction_id(s), then calls annotate_transaction with a note or tag. Server stores and returns the annotated entry.
Monthly spending summary
- Prompt: “Summarize my spending last month by category.”
- Flow: list_transactions for date range, server returns raw transactions; LLM aggregates and returns a summary breakdown.

Security and Best Practices

Treat Monzo credentials and refresh tokens as sensitive secrets; store them in a secrets manager.
Limit server exposure (use API keys, network rules, and TLS).
Audit which scopes and accounts are granted to the server. For shared or joint accounts ensure consent and appropriate access control.
Log

Monzo MCP Server: Manage Balances, Pots, Transactions