Quicken Mac Read-Only Query Skill
Lets Claude query your Quicken for Mac SQLite database in strict read-only mode to answer questions about accounts, spending, and investments.
Evidence shows strict read-only design (sqlite3 mode=ro, no writes/PRAGMAs), parameterized SQL, prompt-injection defense (DB values treated as untrusted), minimized column/row exposure, and refusal to silently pick among multiple databases. Deductions: no executed verification; auto-discovery under ~/Documents is a mild overreach surface; optional snapshot writing path is not enforced within the skill itself.
Probe/user-tag-schema scripts are self-consistent with concrete, actionable error messages (encrypted file, multiple candidates, missing tables); SQL recipes are internally coherent (Core Data epoch, half-open intervals, split-summing rules). Deductions: static review only; the skill's own SQL key paths (ZFISTATEMENT, budget date mapping) have no committed test evidence—repo tests cover the MCP server, not the skill recipes.
SKILL.md states explicit triggers and non-fit scope (generic advice, non-Quicken data) with clear sub-scenarios (balances/cash-flow/investments/budgets/tags). Deductions: boundaries depend on schema that 'can differ by Quicken version'; no Chinese-language support; macOS+Quicken prerequisites are declared but without a compatibility matrix.
Layered docs (SKILL.md→references→optional docs/schema.md), MIT license, SECURITY.md, ROADMAP, version-sync and release workflow, contributor attribution all present. Deductions: some critical details (ZEFFECTIVEDATENUM mapping) are explicitly left to per-file inspection; no skill-level changelog; maintenance concentrated in a single unverified publisher.
Design delivers directly copyable parameterized SQL with a reporting protocol (source, as-of date, currency, transfer treatment), giving clear marginal value over manually reverse-engineering Core Data. Deductions: static review cannot confirm query correctness/completeness on real files; several recipes require per-file validation, so outputs still need checking.
Repo contains CI, test scripts, npm provenance/Sigstore, MCP Registry publishing, and cross-referenced contributor evidence—ample auditable primary material. Deductions: this evidence mostly covers the MCP server; the skill's SQL key paths lack independently reproducible third-party execution evidence, and coverage remains thin under the static cap.
- This is a static source-only review; no queries or scripts were executed, and the skill's SQL recipes are not independently verified against real Quicken files.
- The skill's own key SQL paths lack committed test evidence; repo CI/tests primarily cover the MCP server, not this SKILL.md's recipes.
- Schema identifiers (Z_ENT, join tables) can vary across Quicken versions/files; per-file discovery is mandatory and example values must not be assumed.
- macOS and Quicken for Mac with the app open/unlocked are hard prerequisites; documentation is English-only with no Chinese support.
- Publisher identity is not verified by the FollowSkills registry; verify the repo and npm provenance independently before use.
- The skill touches highly sensitive financial data: trial it in an isolated environment and keep account numbers and similar values out of conversation output.
What does this skill do, and when should you use it?
This is a Claude Skill plus MCP server for Quicken for Mac users. Quicken stores data in a Core Data SQLite database that stays encrypted unless the app is running; the skill teaches Claude to query that database directly with read-only SQL, with no export or modification. The repo also ships an MCP server with eight prepackaged tools, a standalone qmac CLI, and a Python script that exports your full financial history to CSV. macOS only, and Quicken must be open while in use.
Opens the data file inside a Quicken .quicken bundle strictly read-only (sqlite3 -readonly or mode=ro), with a bundled probe script that verifies the database is unlocked and has the expected schema. It then runs SQL to answer questions about accounts, balances, net worth, category spending, monthly trends, payees, transfers, budgets, tags, and investment holdings; it handles Core Data epoch conversion (+978307200), split-amount summation, and half-open date ranges. The MCP server exposes eight tools such as list_accounts, query_transactions, and spending_by_category, and export_sovereign_csv.py dumps everything to normalized CSV plus schema.. It never writes, migrates, or modifies the database.
- A Quicken for Mac power user who wants Claude to answer 'what did I spend on groceries last month?' without manual report exports
- Monthly personal-finance reviews in Claude Code: spending by category, monthly trends, cash flow
- Users of non-Claude MCP clients (Cursor, Cline) who prefer typed tools over raw SQL
- Anyone migrating their complete financial history out of Quicken into DuckDB, Pandas, or PostgreSQL for self-hosted analysis
- Investors checking holdings composition, cost basis, and stored Quicken quotes (holdings analysis only; balances come from statement snapshots)
What are this skill's strengths and limitations?
- Strictly read-only: no writes, migrations, or repairs; parameterized SQL with no string splicing
- Detailed schema knowledge and reference docs covering Core Data timestamps, split amounts, and transfer filtering pitfalls
- Three usage paths: Claude skill, MCP server (8 tools), and standalone qmac CLI
- Zero-dependency Python exporter packages your full history as CSV + schema. for sovereign analysis
- MIT licensed, published on npm and the MCP Registry
- macOS and Quicken for Mac only (not Quicken Windows), and Quicken must stay open or the database is unreadable
- Community project, not affiliated with Intuit; Quicken updates could change the schema
- Stored investment quotes may be stale — the skill itself requires source and as-of dating on every balance
- README provides no evidence of test coverage or compatibility testing across Quicken versions
How do you install this skill?
Recommended (Claude Code, skill + MCP together): run claude plugin install quicken-mac-mcp. MCP-only: claude mcp add quicken -- npx -y quicken-mac-mcp, or add the corresponding mcpServers entry to claude_desktop_config. (or drag the .mcpb from the latest GitHub release into Claude Desktop). Set the QUICKEN_DB_PATH environment variable if you have multiple files or a non-default location. Prerequisites: macOS with Quicken for Mac installed and open.
How do you use this skill?
Make sure Quicken for Mac is open and logged in (the database is encrypted while the app is closed), then ask Claude questions such as 'List my accounts', 'Show my spending by category for 2024', or 'Find all transactions from Costco over $100'. The skill probes and verifies the database first, then runs read-only SQL and reports its assumptions (date range, account scope, currency). Terminal users can also use the qmac CLI (e.g. qmac spending-by-category --start-date 2025-01-01 --end-date 2025-03-31) or run python3 scripts/export_sovereign_csv.py for a full export. See the repo README and SKILL.md for full details.
How does this skill compare with similar options?
No direct competitors are named in the source. Within this repo, the README recommends the skill first when your client supports skills (direct SQL, no MCP server process); use the MCP wrapper only for clients that cannot load skills. Both share the same schema knowledge.