create_shadow_client
ChatGPTCreates a hidden ("shadow") client on this advisor’s book from a first and last name (and an optional email). This is the complete action for "create a client": the client is created hidden and does not appear in get_clients or find_client. Creating a client and adding it to the book are separate actions — promote_shadow_client is the one that makes the client visible, and it applies when the advisor asks for that in its own right rather than as a follow-on to this. Returns the new client’s wid, which is what promote_shadow_client takes as its argument. If the advisor is on more than one team, nothing is created and the result is a needs_team_selection status listing their teams; a second call carrying one of those teamId values creates the client.
find_client
ChatGPTResolves a client, either by an exact record identifier or by a NAME. Given the identifier the client book returns when a call asks for the underlying record identifiers, reads that one client’s own file directly: their names, date of birth, citizenship, relationship, marital status, state of residence, the special-needs, children and pets indicators, the file notes, and whether they are deceased. Who advises them is not included. Given anything else — a household or person NAME the advisor said — searches for the client instead, across every name their client book holds: name, nickname, alias, and family and legal name. A name search answers with names, state of residence, current partner, household relationship and whether the client is deceased — best interpretation first — and each match’s own record identifier, which you give straight back to this tool as the client reference on another call to read that one client’s whole file. So searching by name and then reading the file is two calls to this tool and nothing else. A NAME CAN MATCH MORE THAN ONE CLIENT, and when it does the answer says so: isAmbiguous is true, every candidate is listed with what tells it apart from the others, and NONE of them has been chosen. Ask the advisor which one they meant and wait for their answer before reading any client’s file — picking one yourself shows the advisor the wrong client’s information without either of you knowing. Reads and searches only this advisor’s own clients; there is no way to ask about another advisor’s.
get_balance_sheet
ChatGPTThe client’s ownership balance sheet: their total net worth, and a breakdown by owner — each person, trust, or entity the client’s assets are attributed to, how much of the value each holds, and the assets behind it (what it is, its category, where it is held, and this owner’s share of its value). A jointly-held asset counts once toward net worth but appears in full under each joint owner, so the per-owner amounts can add up to more than net worth, which the stated total gives directly. Takes the client’s record identifier (wid), from get_clients or find_client. Only clients on this advisor’s own book can be read.
get_clients
ChatGPTThe advisor’s own client book. Start here to see who a request is about: each entry carries the client’s names, date of birth, citizenship, relationship, marital status, state of residence, the special-needs, children and pets indicators, the file notes, and whether the client is deceased. Who advises each client is not included — that is not on any surface. Ask about one named client for that client’s own file. NAME MATCHING: match a household name the advisor said against every name each entry says the household is known by, case-insensitively and ignoring the words The, Family, and Household — so “the Brady family” matches a client named Brady. If nothing here matches, resolve the name against the tax-side household names before concluding the client does not exist: a household shown on a tax page can be filed in this book under another surname. Returns only this advisor’s clients; there is no way to ask about another advisor’s.
get_estate_risks
ChatGPTWhat was observed about a client’s estate documents: the gaps, the loose ends, and the things an advisor may want to raise, as the client file recorded them, with the page of the document each one came from. Use it for "what should we look at in this estate plan" or "what is missing". These are observations, not a legal review of the client’s plan, and not advice; the client file re-examines a plan as documents change, so the observations may vary over time and this answer may differ from an earlier one for the same client. An empty result means nothing was observed, which is not the same as the plan having no gaps. Takes the client’s record identifier (wid), from get_clients or find_client. Only clients on this advisor’s own book can be read.
get_executive_summary
ChatGPTThe executive summary of one estate document — a trust, a will, or another named document — as the client file holds it. Read the document inventory first, with the underlying record identifiers included, to find the document and the client it belongs to. Takes the document’s own identifier (vaultId) and the client’s wid. Returns a short summary by default; pass the full form for the longer summary verified by Wealth.com. Each statement in the summary carries the page of the document it came from, so it can be checked against the source. Registration numbers are not part of the answer.
get_federal_tax_constants
ChatGPTThe federal tax law for a year: the ordinary-income brackets, the standard deduction, the long-term capital-gains breakpoints, the AMT exemption and rates, the net investment income tax threshold, the section 199A threshold, the state-and-local-tax deduction cap, the child tax credit, the Social Security wage base, and the Medicare IRMAA income tiers — each given for every filing status. This is TAX LAW, not a client’s return: it is about a year, not a person, and no client is named or read. Use it to answer what the law is, to size a Roth conversion against a bracket, or to compare one year with the next. Only years with a published schedule can be asked for; the years available are the ones listed on the tax-year argument.
get_flowchart
ChatGPTDraw the client’s estate structure as a diagram: the people, the trusts, the charities and the entities in the plan, and how they connect, laid out as the estate plan itself is drawn in a Wealth report. Use it when an advisor wants to SEE the structure — "show me how this estate is set up", "draw the disposition", "what does this trust look like". The diagram appears beside this answer; describe what it shows rather than restating it in full. Takes the client’s record identifier and the document’s — read the client book and the document inventory first. When no diagram can be drawn the answer says so plainly rather than showing an empty one.
get_lookback_report
ChatGPTThe Lookback Report for a client’s filed tax year — the same report the Tax Planning report page shows: filing status and filers, adjusted gross income, taxable income, total income, federal income tax and total tax, capital gains, dividends and deductions, the federal rate summary and bracket table, the income overview, what the tax is made up of, itemized deductions, the business, capital-gains, property and farm schedules, and the state figures — state name, how that state taxes income, state income tax, state taxable income, the state bracket table and the state return summary. It is a report on a filed year, not the text of a document. Every effective rate comes back with its tax dollars and the income amount behind it. Applies to tax returns only — not to an estate plan, a trust, a will, or any document that is not a tax return, which have an executive summary instead; asking for one here returns nothing. Only the client is required. Give a tax year when the advisor names one; otherwise the most recent filed year with a report is used. Name a specific return only when the advisor is asking about that one return.
get_state_tax_constants
ChatGPTA state’s income tax law for a year: the income bands and rates for every filing status, and whether the state taxes income progressively, at a flat rate, or not at all. Like the federal version, this is tax law rather than anybody’s return — no client is named or read — so it also answers relocation questions: ask for two states in the same year to compare them. Only years with a published schedule can be asked for; the years available are the ones listed on the tax-year argument.
get_tax_calculation_inputs
ChatGPTWhat a projection for an unfiled year still needs, and what is already known. Ask it before running the numbers and again as figures arrive: it reports both whether the calculation can run at all and how much of the picture is still missing, which are different questions. It saves nothing and needs no client on file.
list_documents
ChatGPTThe client’s document inventory: one entry per uploaded document, with its name (as a link to that document in Wealth.com), the client file’s own label for what kind it is, when it was uploaded, the date the document itself carries, and its reading status (awaiting upload, still being read, ready to summarise, or unavailable). Each entry also carries whether it is a filed tax return — the only kind of document a lookback report exists for — derived from the underlying record rather than the Type label. A document’s executive summary can be read once the entries are fetched with the underlying record identifiers included. Takes the client’s record identifier (wid), from get_clients or find_client. Only clients on this advisor’s own book can be read.
model_tax_scenario
ChatGPTTry a change against a year the household HAS filed, and see what it would do. Reads the filed return, applies the changes on top, and reports the tax both ways with the difference between them. Amounts are how much to MOVE a figure by, not the new total. TWO YEARS, AND THEY ARE DIFFERENT THINGS: baselineYear is the filed year the figures come from, and targetYear is the year whose tax law prices them. Both are required, and a filed year before 2025 cannot be priced under its own law at all, so targetYear may differ from the filed year. Effective rates are returned together with the tax amount and income basis behind them. It saves nothing; use save_tax_scenario to persist a scenario.
promote_shadow_client
ChatGPTAdds a hidden ("shadow") client to the advisor’s book — clears its hidden tag so it appears in get_clients and find_client. It applies when the advisor asks to add, reveal, or make visible a client that create_shadow_client made earlier; creating a client does not imply adding it to the book, so this is a request of its own rather than a follow-on to that one. Takes the wid create_shadow_client returned. A wid that is not one of this advisor’s hidden clients is refused.
run_tax_calculation
ChatGPTThe tax on a year the household has not filed, worked from figures you supply rather than from a return on file. Give the year, how they file, and whatever is known; it returns the federal (and optionally state) tax, or reports that it cannot and what would let it. Every figure comes back attributed to where it came from. Effective rates are returned together with the tax amount and income basis behind them. A position does not say whose account paid out, so on a married_filing_jointly position the 10% additional tax on an early distribution — charged on ira_distributions_taxable or traditional_ira_distribution_taxable and on pensions_annuities_taxable — is owed unless BOTH filers are 59½ or older; with one filer on each side of 59½ the total turns on whose account paid out, and for an owner who reaches 59½ during the year, on whether it is paid before or after. Single, head_of_household and married_filing_separately positions use the primary filer’s age; a qualifying_surviving_spouse position is exempt when either age is 59½ or older. An amount paid from the account of a filer who is 59½ or older when it comes out goes in as the distribution AND again, the same amount, as exempt_ira_distributions (exempt_pension_distributions for a pension or 401(k)), which takes the 10% off that amount and adds no income of its own. roth_conversion_taxable is already exempt: the calculation adds it to exempt_ira_distributions itself, so a conversion also given there is counted twice and can shelter an ordinary withdrawal. That 10% is counted INSIDE totalTax, with no figure of its own. Nothing is saved to the client’s file.
find_client
Claudeget_clients
Claudeget_estate_risks
Claudeget_estate_snapshot
Claudeget_executive_summary
Claudeget_lookback_report
Claudelist_documents
Claude