finlynq

Finlynq is an open-source (AGPL v3) personal finance management application with a built-in first-party MCP server, letting you track your finances here and analyze them anywhere.

Purpose: For personal finance management, supporting recording and querying of income, expenses, budgets, investments, loans, subscriptions, and goals. Through the MCP server, you can query and manage financial data in Claude, Cursor, or any MCP client.

Key capabilities:

• 100+ MCP tools (HTTP and stdio): covering accounts, transactions, budgets, goals, loans, portfolios, subscriptions, exchange rates, rules, splits, bulk edits, and file imports. Of these, 55 are HTTP and 89 are stdio.
• Encryption: AES-256-GCM envelope encryption, scrypt-derived KEK, per-user keys encrypting sensitive fields at rest.
• Import: supports CSV, Excel, OFX/QFX, and PDF import.
• Financial management: budgets, portfolios, goals, loan management.
• AI chat: natural language AI chat.
• Calculation tools: FIRE calculator and Monte Carlo simulation.
• Automation: rules and auto-categorization.
• Deployment: self-hosted or managed cloud.
• Interfaces: REST API + MCP (HTTP and stdio, OAuth 2.1 + DCR).
• Interface: dark mode, mobile-friendly.

How to use: Two ways to use it: managed cloud or self-hosted.

1. Managed cloud: visit finlynq.com/cloud, sign up, and import a CSV, with no infrastructure to manage. A public demo is at finlynq.com/cloud?demo=1 (credentials pre-filled, reset nightly). To connect Claude, paste https://finlynq.com/mcp into Claude → Customize → Connectors.
2. Self-hosted: use Docker. Before the first launch, create a sibling .env file containing three secrets (each generated with openssl rand -hex 32): PF_JWT_SECRET, PF_PEPPER (≥32 characters), PF_STAGING_KEY (≥32 characters). If any is missing, Compose fails fast with a clear message. Then open http://localhost:3000 and sign up. The app and PostgreSQL run in Docker; sensitive fields are encrypted at rest using per-user keys derived from the account password. Before exposing the container beyond localhost, change the default PostgreSQL password (POSTGRES_PASSWORD in .env). Full setup instructions are at finlynq.com/self-hosted.

MCP client connection:

• Claude Web / Mobile / Cursor / Windsurf: use OAuth 2.1 + dynamic client registration. Paste https://finlynq.com/mcp into the connector settings, with no config file needed.
• Claude Desktop (stdio): point to mcp-server/index.ts and set PF_USER_ID in the environment block.
• Bearer API keys: generate pf_* tokens in Settings → API Keys. Works for the MCP endpoint and REST routes under /api/. Note that pf_* keys are unscoped and do not expire until regenerated, and should be treated as full access to the account; the REST interface is the interface used by the web and mobile apps, and its response structure is not guaranteed to be stable—MCP is the supported integration interface. Submitted to the Anthropic Connectors Directory (2026-05-09). Full client setup, tool catalog, and brokerage statement recipes are in the Connect Your AI guide.

Use cases: Personal finance management, users who want to query and manage financial data from AI clients, and users who want to self-host to protect data privacy.

Limitations: AGPL v3 license. If you run a modified version as a network service, you need to provide the source code to users. The managed service finlynq.com/cloud runs the same code as this repository. If you want to build a product on Finlynq that cannot bear that obligation, you can obtain a commercial license, see COMMERCIAL.md. The project is donation-based, with no paid tiers; if Finlynq is useful to you, you can support it via GitHub Sponsors or Ko-fi.