Financial Planning Rule Sets
Cited, current, callable.
Every January the IRS moves the numbers your software runs on, and the constants you hard-coded last year quietly go stale. Rule Sets keeps them current: every U.S. tax, benefit, and planning figure traced to its primary government source and served over a live API and MCP, so you pull the right value at request time instead of re-keying it and hoping.
GET /api/reference/v1/us-federal-2026 { "key": "retirement.401k.elective_deferral.under_50", "value": 24500, "unit": "USD", "tax_year": 2026, "effective_date": "2026-01-01", "status": "official_final", "source": { "issuer": "IRS", "document": "Notice 2025-67", "section": "402(g)(1)" } }
Every figure carries its issuer, document, section & effective date.
Federal, plus five states, and growing by demand.
103 source-verified federal figures for tax year 2026, plus state planning parameters for the cases that move the needle. Each value cites the IRS, SSA, CMS, HHS, or Treasury document it came from.
A live API and an MCP server. No scraping.
Call the JSON API directly, or point Claude, Cursor, or your own app at one MCP endpoint and let it look up a cited figure on demand. Either way, the source comes back attached, so the assistant you ship can show its work too.
Versioned JSON API
GET /api/reference/v1/us-federal-2026 returns figures, units, effective dates, and citations. ETag-cached.
MCP tools, agent-callable
get_reference_parameter, list_reference_parameters, get_planning_benchmark.
Point-in-time history
Ask what a figure was as of a past date, and defend a number you shipped last quarter.
// add the server once, for Claude, Cursor, or your app { "mcpServers": { "wealthschema": { "url": "https://www.wealthschema.com/api/mcp" } } }
Cited, not guessed.
Every figure is compiled from its primary U.S. government source (issuer, document, section, URL, and effective date) and cross-checked before it ships. “IRS Notice 2025-67,” not “a model told me.” Golden-fixture regression tests mean a new tax year cannot silently change the shape or break a key.
What the feed fixes, measured.
In a pre-registered benchmark, bare frontier models cited a stale or fabricated regulatory figure in 29% of figure-bearing attempts. The same tasks re-run with this feed's figures in context: zero stale figures in 225 attempts across three systems — and the IRMAA boundary task every bare system failed passed on every attempt.
| System | Bare Stale Figure Rate | With Rule Sets feed |
|---|---|---|
| claude-sonnet | 0.07 (5/75) | 0.00 (0/75) |
| gemini-3.1-pro-preview | 0.17 (13/75) | 0.00 (0/75) |
| claude-haiku | 0.63 (47/75) | 0.00 (0/75) |
k=3, 25 figure-bearing tasks, protocol pre-registered; full tables, substitution log, and caveats on the benchmark scoreboard and methodology page. Measure your own system: AI Eval Sets.
See the sample free. Subscribe for every figure.
The free sample shows the format and the citations on 10 headline figures. Every paid plan includes all federal and state figures; higher plans add history, webhooks, redistribution, and a freshness SLA.
10 headline federal figures, so you can see the format and citations. No card.
Every figure: federal + 5 states. Single-product commercial use, monthly rates, MCP.
Point-in-time history, update webhooks, multi-environment.
Redistribution within your product, freshness SLA, priority support.
Planning-software vendors, custodians & large RIA tech.
See the full rights & terms on the feed terms page.
Informational only — not tax, legal, or financial advice. Each figure is compiled from the cited primary U.S. government source and cross-checked. The 2026 feed is a preview; confirm any value against the issuing agency before relying on it.
Get the 2027 figures the week they are published.
The IRS, SSA and CMS release next year's limits, brackets, COLA and IRMAA tiers over October and November. We'll email you when each set is live in the feed, with its source citation. Release notices only; unsubscribe with one click.
Track the 2027 releasesFrequently asked questions
What does Rule Sets provide?+
Cited, current U.S. financial-planning figures from the IRS, SSA, CMS, and Treasury, including contribution limits, tax brackets, RMD divisors, IRMAA tiers, COLA, and estate and gift numbers. Each value carries its issuer, document, section, and effective date, and is served over a JSON API and an MCP server.
Is there a free tier?+
There's a free sample: 10 headline federal figures, no account or card needed, so you can see the format and the citations. Every figure, federal and state, is included in paid plans from $49/month; higher plans add point-in-time history, update webhooks, and redistribution rights. Because the figures change with the tax code, keeping them current is a service, so Rule Sets is a subscription.
How is this different from the synthetic Data Sets?+
Rule Sets is a living reference feed and a subscription because the figures change over time. The Wealth Data Sets and DecisionSynth are static, so they are one-time purchases you own. Rent what changes; own what does not.
Can my AI agent use it over MCP?+
Yes. Point Claude, Cursor, or your own application at one MCP endpoint and the agent can look up a cited figure on demand with its source attached, so the assistant you ship can show its work instead of guessing.
Which jurisdictions are covered?+
U.S. federal plus five states today, with more states routed by paying-customer demand. Tell us which two you need and they move up the queue.
Are the figures defensible for compliance?+
Yes. Every value cites its primary government source with issuer, document, section, and effective date, and passes golden-fixture regression tests so a new tax year cannot silently change a key. That is the difference between a number you can defend in an audit and one you hope is right.