mcpField
L
Local-first personal finance MCP

Local-first personal finance MCP

Local
github.com/adelaidasofiaFinance0Updated 1d ago

About Local-first personal finance MCP

finance-mcp is a local-first personal finance MCP for people who want to keep account data in a local markdown vault. It uses Plaid to sync balances, holdings, transactions, and liabilities, while storing access tokens in macOS Keychain.

Key features

  1. 01Link bank and brokerage accounts through Plaid
  2. 02Sync balances, holdings, transactions, and liabilities
  3. 03Write account and cash-flow data to Markdown
  4. 04Store access tokens in macOS Keychain
  5. 05Keep transaction history in local SQLite

Use cases

  1. 01Track balances across linked financial accounts
  2. 02Review investment holdings and monthly cash flow in a vault
  3. 03Sync transactions locally without returning access tokens

MCP support

The MCP server exposes tools to link and unlink institutions, sync financial data, roll up monthly cash flow, and inspect audit logs. Tool responses reference institutions by alias and do not include raw access tokens.

Configuration

Local server config is the same for every client. Copy it into your client’s MCP config file.

No install method listed yet

We haven’t collected a package or remote endpoint for this server. Follow the instructions in the project README.

Repository
Packages
mcpbfinance-mcp.mcpb0.1.0

Tools

No tool list yet

Local servers expose tools only once running on your machine, and the maintainer has not provided a tool list.

Details

From the project README

License GitHub stars Last commit Open issues PyPI version PyPI downloads Built by Mycelium AI

finance-mcp

Local-first personal finance MCP. Aggregates bank, brokerage, credit, and loan accounts via Plaid. Writes balances, holdings, and transactions to your local Obsidian-style markdown vault. Access tokens stay in macOS Keychain. Never sends data anywhere except plaid.com.

What it does
  • Link any US bank or brokerage via Plaid (Chase, Citi, Schwab, Discover, Fidelity, Wells Fargo, Bank of America, Capital One, Vanguard, Robinhood, Coinbase, student-loan servicers, 12,000+ institutions total).
  • Pull live balances for every account → writes the auto-sync block in your vault's Accounts.md.
  • Pull investment holdings with shares, cost basis, current value → writes Investments.md.
  • Pull transactions cursor-style (delta only after first sync) → stores in local SQLite.
  • Roll up monthly cash flow by category → writes Cash Flow.md.
  • Pull liabilities (credit-card APRs, statement balances, student-loan payoff info).
  • Audit log every Plaid call and every keychain access.
Why local-first matters

Most personal-finance SaaS (Mint, Copilot, YNAB, Monarch) puts your bank data on their servers and charges you for the privilege. This MCP:

  • Runs on your laptop, talks directly to Plaid, writes to your local vault.
  • Stores access tokens in macOS Keychain, not env files or any database.
  • Never returns access tokens in tool responses — items are referenced by alias (chase, citi, etc.).
  • Vault data is plain markdown with inline Dataview fields — your data, your format, queryable forever.

Install

Open Claude Code, paste:

/plugin marketplace add adelaidasofia/finance-mcp
/plugin install finance-mcp@finance-mcp

Requires macOS (for Keychain) and Python 3.11+. After install, set your Plaid credentials in .env at the plugin root (see SETUP.md for the Plaid signup walkthrough).

First-time use
> healthcheck

If it shows blockers, follow them. Once green:

> link_start(institution_alias="chase")

Open the returned link_token in Plaid's Link demo page (paste the token in the field labeled "Link Token", then click "Open Plaid Link"). Complete bank auth in the browser. Copy the public_token from the success page.

> link_complete(institution_alias="chase", public_token="public-sandbox-...")

Repeat for each bank.

> sync_balances
> sync_holdings
> sync_transactions
> rollup_month(month="2026-05")

Your vault's Accounts.md, Investments.md, and Cash Flow.md now have auto-sync blocks with current data. Re-run any time.

Tool surface
ToolWhat it does
healthcheckVerify Plaid creds, vault path, keychain access.
link_start(alias)Start linking a new bank. Returns link_token.
link_complete(alias, public_token)Exchange public_token, store in Keychain.
list_linkedList linked institutions + last-sync timestamps.
unlink(alias)Remove an institution. Revokes Plaid item + deletes Keychain entry.
sync_balances([alias])Pull current balances → Accounts.md.
sync_holdings([alias])Pull investment positions → Investments.md.
sync_transactions([alias])Cursor-based transaction sync → SQLite.
sync_liabilities([alias])Pull credit + loan details.
sync_all([alias])All of the above in sequence.
rollup_month(month)Compute monthly income/expense rollup → Cash Flow.md.
audit_tail(n)Last N audit-log entries.
Security model
  • Plaid access_token lives in macOS Keychain (security CLI), service name finance-mcp, account name = your institution alias.
  • Plaid client_id + secret live in .env (chmod 600). Never committed (in .gitignore).
  • Transaction history lives in SQLite at ~/.claude/finance-mcp/data.db (not in your vault).
  • Vault writes only happen inside the configured FINANCE_MCP_FINANCE_FOLDER. Set to empty string to disable vault writes entirely.
  • Audit log records every Plaid call and every keychain operation.
  • Tool responses never include raw access tokens.
Plaid environments
  • sandbox (default): fake banks, fake credentials (user_good / pass_good). Free forever. Use this first.
  • development: real banks, free up to 100 items per Plaid account. Use this for personal accounts.
  • production: real banks at scale. Requires a Plaid application + paid plan.

Set via PLAID_ENV in .env. Switch by re-linking all institutions (tokens are environment-bound).

Companion MCPs
  • jkoelker/schwab-mcp — Direct Charles Schwab Trader API for options, trade placement, and deep position data beyond what Plaid exposes.
  • tomasgesino/schwab-mcp — Schwab wheel-strategy management with dry-run-default trading.

finance-mcp covers Schwab basic balances and holdings via Plaid; the above are the route for Schwab power-user features.

Related MCPs in this family

Telemetry

This plugin sends a single anonymous install signal to myceliumai.co the first time it loads in a Claude Code session on a given machine.

What is sent:

  • Plugin name (e.g. slack-mcp)
  • Plugin version (e.g. 0.1.0)

What is NOT sent:

  • No user identifiers, names, emails, tokens, or API keys
  • No file paths, message content, or anything from your work
  • No IP address is stored after dedup processing

Why: Helps the maintainer know which plugins people actually install, so attention goes to the ones that get used.

Opt out: Set the environment variable MYCELIUM_NO_PING=1 before launching Claude Code. The hook will skip the network call entirely. Already-pinged installs leave a sentinel at ~/.mycelium/onboarded-<plugin> — delete it if you want to reset state.

License

MIT. See LICENSE.


Legacy install (manual)

For users who can't use the plugin marketplace yet, the manual flow:

git clone https://github.com/adelaidasofia/finance-mcp ~/.claude/finance-mcp
cd ~/.claude/finance-mcp
pip3 install --break-system-packages -r requirements.txt
cp .env.example .env
chmod 600 .env

Then register with Claude Code by adding to your vault's .mcp.json (project scope) or ~/.claude.json (user scope, via claude mcp add):

"finance": {
  "type": "stdio",
  "command": "python3",
  "args": ["-m", "finance_mcp.server"],
  "env": {
    "PYTHONPATH": "/Users/<you>/.claude/finance-mcp",
    "FINANCE_MCP_VAULT_PATH": "/Users/<you>/Documents/MyVault",
    "FINANCE_MCP_FINANCE_FOLDER": "Finance"
  }
}

Restart Claude Code. Tools appear under mcp__finance__*.


Built by Adelaida Diaz-Roa. Full install or team version at diazroa.com.

Comments

Markdown supported

New to MCP?

From what the protocol is to installing your first server, these answers cover most first questions.

An open protocol that defines how AI applications connect to external tools and data in a uniform way. Clients talk to the model; MCP servers provide tools, resources and prompt templates. They exchange standard messages, so one server can be reused by many clients.