The metric standard your AI agents trust.
Regulated metric definitions wired into MCP. Every formula carries its source paragraph, its lineage, and a signed audit trail — so when your agent quotes CET1 ratio, it's the same number the regulator expects.
Deploy custom metrics alongside the regulatory standard
Define your own metrics in the custom.<org> namespace. They live in your private workspace, resolve over the same API, and never leave your tenant.
Define privately
Write metric YAML in the custom.<org>.* namespace. Your definitions are private to your tenant — other users never see them.
Resolve over the same API
Custom metrics are served via the same REST and MCP endpoints as regulatory metrics. One integration, both sources.
Promote when ready
When a custom metric is battle-tested, export it to the public marketplace for community review and certification.
id: custom.acmebank.treasury.net_interest_margin
title: Net Interest Margin (Internal)
industry: banking
formula:
dialect: ansi
sql: |
SELECT (interest_income - interest_expense)
/ NULLIF(avg_earning_assets, 0) AS nim
FROM {{treasury_daily}}
WHERE report_date = :as_ofThree reasons your AI agent gets the number wrong.
An LLM asked “what’s our gross margin?” will confidently invent the formula. There’s no grounding source.
When a board member queries a number, you can’t tell them which dataset, which formula version, or which paragraph of regulation it traces to.
Three teams, four spreadsheets, six dashboards, all subtly disagreeing on the same KPI. The agent picks one at random.
Crawl → Extract → Serve → Audit.
Regulatory source monitoring — starting with CRR, PRA Rulebook, and Basel III. More sources added as industries launch.
LLM-assisted extraction normalises every metric into formula + inputs + outputs + paragraph reference.
Any MCP-capable client (Claude, Cursor, your gateway) resolves a metric ID to its canonical formula in one call.
Hash-chained log of who resolved what, when, against which version. Exportable for regulators.
Three industries. One contract.
Capital, liquidity, credit risk, conduct.
Outcomes, safety, throughput, quality.
Climate, social, governance disclosures.
Eight things you only get here.
- Paragraph-level regulatory citation
- Hash-chained audit log
- Version pinning + diff history
- Validator marketplace
- MCP-native resolution
- SQL canonical form
- Industry-specific schemas
- CC BY 4.0 licensing
# banking.capital.cet1_ratio id: banking.capital.cet1_ratio title: CET1 Ratio version: 3.2.0 formula: | {{cet1_capital}} / {{rwa_total}} source: framework: Basel III paragraph: CRR Art. 92(1)(a) effective: 2026-01-01 unit: percent frequency: monthly
Open. Peer-reviewed. Transparent.
Anyone with the credentials can submit a metric definition for review. No gatekeeping.
Two domain validators sign off before a metric goes from draft to certified.
Every approval and rejection ships with a public reason. The reputation graph is on-chain.
Ready to give your AI agents a single source of truth?
Start with the open tier — no credit card. Wire MCP in three lines.