wealthdb
Personal Wealth Management Suite
Install / Use
claude mcp add smartpointer-com -- npx -y github:smartpointer-com/wealthdbIf the server publishes to npm under a different name, use that package instead — check the repo README.
MCP Server
Model Context Protocol server
Quality Score
Category
AutomationSupported Platforms
Our assessment of wealthdb
wealthdb scores 76/100 on our quality scale, 2590th of 2,853 Automation skills we index.
Its MCP Server is 20 KB long, well organised into 11 sections with 5 code examples: a thorough specification that gives an agent plenty to work with.
It has 3 GitHub stars, so there is little community track record yet; judge it on its content.
Maintenance, license and trust
- The repository was last updated today, so wealthdb is actively maintained.
- It is released under the MIT license, a permissive license that allows use, modification and commercial use with attribution.
- Its trust signals score 87/100, with 2 cautions from licensing, adoption, age or documentation. These come from repository metadata, not a code audit — read the skill file before letting an agent act on it.
wealthdb compared with similar skills
All 4 of these similar skills score higher than wealthdb; compare them before choosing.
| Skill | Score | Stars | Updated | Format |
|---|---|---|---|---|
| wealthdb (this skill)by smartpointer-com | 76 | 3 | today | MCP Server |
| claude-memby thedotmack | 100 | 112.1k | today | CLAUDE.md |
| Agent-Reachby Panniantong | 100 | 108.5k | 3d ago | CLAUDE.md |
| headroomby headroomlabs-ai | 100 | 78.8k | today | CLAUDE.md |
| CowAgentby zhayujie | 100 | 47.8k | today | CLAUDE.md |
Frequently asked questions
- How do I install wealthdb?
- Run
claude mcp add smartpointer-com -- npx -y github:smartpointer-com/wealthdb. The install tabs above show the steps for each supported agent. - Which AI agents does wealthdb work with?
- It is written for Claude Code and Claude Desktop, as a MCP Server file. Other agents that read the same format can often use it too.
- Is wealthdb safe to use?
- It is MIT-licensed and scores 87/100 on trust signals. Skills are instructions an agent will follow, so read the file before installing it and do not approve commands you do not understand.
- Is wealthdb still maintained?
- The repository was last updated today, so wealthdb is actively maintained.
Skill content
View source on GitHubwealthdb
A personal-finance and investment-performance data suite.
It collects holdings, transactions and documents from banks, brokerages, pension providers and private-market platforms. It consolidates them into one local database. From that database, one CLI and an optional set of dashboards answer:
- what was held, anywhere, as of any date;
- what is really owned, every fund looked through to what it holds;
- what the holdings yield, and how they move with interest rates;
- what each account, each portfolio and the whole returned;
- what came in and what went out, typed and categorised;
- where the cash went, as a cash flow statement with the household's own internal moves taken out.
The collectors download, the engine consolidates, the reports read. All data stays on the local machine, and no service sits in between, with three exceptions:
- A login linked through Plaid. Plaid reads that institution and keeps a copy of its data.
- The model that
wealthdb categorizeandresolve-symbolsask. When it runs on a remote endpoint, it receives the merchant, payer and instrument names it is asked about (SPENDING.md §5). - The
refdatacollector. It asks public sources about each fund and security held, by its identifier. They see which securities are asked about. Each source also sees who asks, FINRA by the API client its requests carry. No amount, account or bank is sent.
⚠️ Security & liability disclaimer
[!WARNING] The wealthdb suite handles fully privileged financial-account credentials. Read this disclaimer in full before configuring any credential anywhere in the suite.
The collectors sign in to banks, brokerages, and pension providers with your credentials and your multi-factor confirmations. The web-scraping collectors impersonate a human browser user (a stealth-hardened browser session), and the API collectors hold write-capable credentials; in both cases the session is fully privileged — the same login a human uses to move money — and no provider offers a read-only sub-scope. Nothing but the codebase's own discipline restricts the collectors to reading. Plaid is the exception: its tokens read data and cannot move money. Plaid's keys and tokens still expose the data of every linked login. If malicious code were ever introduced into this repository, its dependency chain, or the container images it runs, it could act on your accounts with your full authority and cause irreversible financial damage, up to the total loss of the assets reachable from those credentials.
You are solely responsible for a thorough, independent security audit of this code, its dependency chain, and its runtime images before entrusting the suite with credentials, and again after every update or rebuild. If you cannot perform such an audit, do not hand this software real credentials. Automated access may additionally breach a provider's terms of service; verifying that your use is permitted is likewise your responsibility.
No warranty; no liability. This software is provided “AS IS”, without warranty of any kind, express or implied, including but not limited to the implied warranties of merchantability, fitness for a particular purpose, title, and non-infringement. To the maximum extent permitted by applicable law, SmartPointer AG and the contributors accept no responsibility for, and shall not be liable for, any claim, damages, or other liability — whether in an action of contract, tort, or otherwise — arising from, out of, or in connection with this software or its use, including without limitation unauthorized or erroneous transactions, loss of funds or other assets, credential or data compromise, account suspension or termination, and any direct, indirect, incidental, special, consequential, or punitive damages. Your use is entirely at your own risk. See LICENSE for the governing terms. This software is not affiliated with, endorsed by, or sponsored by any financial institution; nothing in this repository is financial, legal, or tax advice.
What it does
wealthdb gives one private, complete view of a household's money, across every account it collects, and answers the questions a household asks of it.
Net worth and allocation, as of any day. What is held, where, and what it is worth, in one currency at that day's rates. Holdings are classified by what they are exposed to and how they are held, so the answer can be "a third listed equity, a fifth real estate, a tenth of it inside retirement plans" rather than a list of tickers.
Investment performance. Time-weighted and money-weighted returns for each account, each portfolio, each provider and the whole, over any period. Money moved between the household's own accounts is not counted as gain or loss, and every figure carries its own caveats. Realized and unrealized gains come from the cost basis and tax lots the providers state. A crypto exchange or an old brokerage statement often states none. There a lot engine rebuilds them from the trades, first in first out by default. Every rebuilt figure says so.
Exposure. What is really owned: every fund looked through to the
stocks and bonds it holds, added to what is held directly, by asset
class, sector, size, country, currency, credit, maturity and issuer,
and where each holding legally sits. The fund data comes from public
sources the refdata collector reads. Where nothing describes a fund,
a proxy the config names stands in, and every gap is listed.
Yield. What the holdings earn, and how their value moves with
interest rates. Each holding gets its yields. Each bond, fund and
deposit also gets its duration and its rate sensitivity. Cash is set
against its currency's benchmark rate. The public yield curves come
from the fred and curves collectors.
Income and spending. What arrived — wages, interest, dividends, rent — and what left, grouped by category, across all cards and accounts, and what the holdings are expected to pay. The household's own transfers are taken out, so a move from checking to savings never looks like spending.
Cash flow. A cash flow statement for the household: what came in, what went out, what was invested or sold, what went into or came out of retirement plans and trusts, and what was left. As a table, and as a Sankey diagram.
A few of the questions it answers, and the commands behind them:
wealthdb holdings global -d 2024-12-31 -x CHF # net worth at the end of 2024, in CHF
wealthdb returns portfolios 2025 # how each portfolio did in 2025
wealthdb gains realized 2025 # every lot sold in 2025, with its gain
wealthdb exposure countries # where the money is, every fund looked through
wealthdb yield global -x CHF # what the holdings yield, by currency
wealthdb spending categories 2025 --period annual # last year's spending by category
wealthdb income summary 2025-01 2025-03 -C +withheld # Q1 income, with the tax withheld beside it
wealthdb cashflow sankey 2025 # the year's cash flow as a diagram
wealthdb cashflow projection --period quarterly # the next twelve months' cash flow projected, by quarter
Every report is available in three places:
- the command line, as a table, CSV or JSON;
- a set of local dashboards (
wealthdb web) for net worth, allocation, returns, gains, spending, income, cash flow, exposure and yield; - AI agents, over MCP (
wealthdb mcp).
A privacy mode hides amounts, for a screen others may see. The CLI's
-p redacts them, and most dashboards have a twin that shows shares
instead.
wealthdb help lists every command with its flags; the topic documents
under Documentation cover each feature in depth.
Use with AI agents
The same reports are safe to put in front of an AI agent. An agent is genuinely useful when it can answer questions over a complete financial picture. It is genuinely dangerous when the way to get there is handing it banking credentials: a fully privileged e-banking login in the hands of a probabilistic, prompt-injectable system is a standing invitation for irreversible damage. wealthdb keeps the two apart:
- Credentials are handled only by static, reviewable code. The collectors are deterministic scripts that can be audited line by line. A person starts each run and answers each MFA challenge. No agent drives a banking session, and no agent ever sees a credential. A login linked through Plaid is handled by Plaid's code instead. Its sign-in starts on a page Plaid hosts, and Plaid holds the access to the institution. The collector keeps Plaid's keys and one token per login. A token reads data and cannot pay.
- The collectors only read. Their contract is navigate, filter, export. No code path submits a form, places an order or changes a setting, and no CLI flag can enable one.
- The data lands locally. Everything is parsed into local
databases and consolidated into one queryable gold store.
Nothing is sent to any third-party service, with the three
exceptions named at the top: Plaid, a remote model that
categorizeandresolve-symbolsask, and the public sourcesrefdataasks about each holding. - Agent access is read-only by construction. The MCP server
(
wealthdb mcp, mcp/README.md) serves the reports as tools. It has no tool that writes, it opens the database read-only for one call at a time, and its container mounts the data read-only. An agent with a shell can run the CLI instead, with--read-only. skills/wealthdb-ro/SKILL.md teaches an agent which commands to run. Either way, an agent sees only consolidated, local, read-only queries.
The result is a clean separation. An agent runs the same reports a
person does, through the MCP server or wealthdb --read-only. The
server's privacy endpoint, or -p on the CLI, keeps amounts out of a
transcript. No agent is ever given write access to the financial data,
let alone the banking credentials that produced it.
Data sources
There is one collector per provider. collectors/README.md has the full table.
- Banks — UBS (netbanking and the PSN feed), Swissquote, Raiffeisen Austria, Chase, First Citizens; American Express cards.
- Brokerages — Schwab (the Trader API and the client web), Fidelity.
- Pensions — VIAC (pillar 3a), Relevate (pillar 2 vested benefits).
- Crypto — CoinTracking.
- Aggregators — Plaid reaches the banks, brokers and card issuers it covers, one linked login at a time.
- Private markets — AngelList, Carta, EquityZen.
- Archives and reference — SVB statement
archives (load-only);
manualfor holdings with no portal, such as property, private loans and escrow claims; FRED for historic FX rates and the Treasury curve;curvesfor the euro and franc government yield curves;refdatafor what each fund holds and states about itself, what each company reports, and the distributions each fund and share pays.
Adding a data source
wealthdb is designed to be extensible. Adding a bank, brokerage or pension provider means writing one coll
Truncated for display — read the full file on GitHub.
Related Skills
claude-mem
112.1kPersistent Context Across Sessions for Every Agent – Captures everything your agent does during sessions, compresses it with AI, and injects relevant context back into future sessions. Works with Claude Code, OpenClaw, Codex, Gemini, Hermes, Copilot, OpenCode + More
Agent-Reach
108.5kGive your AI agent eyes to see the entire internet. Read & search Twitter, Reddit, YouTube, GitHub, Bilibili, XiaoHongShu — one CLI, zero API fees.
headroom
78.8kCompress tool outputs, logs, files, and RAG chunks before they reach the LLM. 20% fewer tokens for coding agents, 60-95% fewer tokens for JSON, same answers. Library, proxy, MCP server.
CowAgent
47.8kOpen-source personal AI assistant & Agent Harness. Plans tasks, runs tools and skills, self-evolves with memory and knowledge. Multi-agent, multi-model, multi-channel. Lightweight, extensible, one-line install.
Languages
Trust signals
From repository metadata: license, adoption, age and documentation. Not a code audit — see the Safety scan above for what the skill file itself contains.
