Risk Calculation MCP
This server does not currently have a deployable release on Glama. Follow the repository's setup instructions to install and configure it for your MCP client, including any required credentials and local dependencies.
Risk Calculation MCP
Free, local credit-risk calculations for MCP clients. IRB and standardised risk weights, risk-weighted assets (RWA), and the 8% capital requirement under EU CRR/CRR3 and Turkish BDDK current/draft rules. Calculations use the same public core as rwacalculator.io. Source: github.com/mselcukozturk/risk-calculation-mcp.
The publisher does not operate a public MCP endpoint. Run it on your computer or your own server. No accounts, API keys, model API, telemetry, storage, or outbound network requests are used during calculation. Your AI application has its own data handling policies; local MCP execution does not mean your conversation stays on your computer.
Install
Requires Node.js 20 or later. Standard MCP configuration for Claude Desktop, Cursor, VS Code and other clients supporting stdio:
{
"mcpServers": {
"risk-calculation": {
"command": "npx",
"args": ["-y", "risk-calculation-mcp@1.2.1"]
}
}
}The first npx run downloads the package and dependencies from npm. Pin the version for
repeatable calculations. To avoid downloads at each client startup, install it once with
npm install -g risk-calculation-mcp@1.2.1 and use risk-calculation-mcp as the command.
If the client cannot find Node or npx, use the full path to the executable.
Claude Code:
claude mcp add --transport stdio risk-calculation -- npx -y risk-calculation-mcp@1.2.1Codex:
codex mcp add risk-calculation -- npx -y risk-calculation-mcp@1.2.1A Claude Desktop .mcpb installer is available from the
installation page. The bundle includes dependencies;
the host application supplies the Node runtime.
Related MCP server: RiskProbe Agent
Tools
get_risk_capabilities: supported rule books, classes, input units and limitations.calculate_irb: single non-defaulted corporate, corporate SME, mortgage, qualifying revolving retail or other retail exposure; effective floors, correlation, RWA and capital.calculate_standardised: 18 exposure classes in modern rules and the supported subset in legacy rules; CCF, rule reference, RWA, capital and informational output floor rate.
Both calculation tools require an explicit regime: crr (EU before 2025), crr3 (EU
from 2025), bddk (current Turkish rules), or bddkDraft (draft, not current law).
This convention is identical across both tools. Language never selects jurisdiction.
PD, LGD, LTV, CCF and risk weights use percentage points: pd: 1 means 1%, not 0.01%.
Amounts are currency units (0–1e15), maturity is years (1–5; corporate default 2.5),
annualSales is millions of EUR for EU or TRY for BDDK. Article 501 smeTotal is always
EUR, separate from exposure currency. No FX conversion is performed.
Under CRR3 and the BDDK draft, calculate_irb floors own LGD estimates at the unsecured
values: 25% corporate, 5% residential mortgage, 50% qualifying revolving retail and 30%
other retail. Set foundation: true for a corporate foundation-approach exposure: pass
the supervisory LGD (40%, 45% or 75% under CRR3, 45% or 75% before) or an LGD* that
already reflects eligible collateral; the own-estimate floors are then not applied and
LGD* is not computed. qrreTransactor: true selects the CRR3 0.05% PD floor for a
QRRE transactor instead of 0.10%. Neither is inferred.
Example call to calculate_irb:
{
"regime": "crr3",
"exposureClass": "corporate",
"pd": 1,
"lgd": 40,
"maturity": 2.5,
"ead": 1000000,
"currency": "EUR"
}Risk weight is approximately 82.06%. Results include supplied and effective inputs, rule-book status, sources, warnings and links to the matching calculator. Monetary results remain null when no amount is given. The 8% capital figure excludes buffers and institution-specific requirements.
Standardised example:
{
"regime": "crr3",
"exposureClass": "corporate",
"rating": "3",
"year": 2026,
"item": "commitment",
"amount": 1000000,
"currency": "EUR"
}Risk weight 75%, CCF 40%, exposure value EUR 400,000, RWA EUR 300,000, capital EUR 24,000.
Inspect the tool schema for class-specific fields. Missing material facts produce a
tool error; preferential eligibility is never inferred. Corporate SME IRB requires
annualSales; BDDK draft additionally requires smeLimit in TRY millions. Current BDDK
uses C=1000 million TRY, EU uses EUR 50 million. Those limits are versioned assumptions,
not live regulatory data.
Your own HTTP server
npx -y risk-calculation-mcp@1.2.1 --httpThis serves stateless Streamable HTTP at http://127.0.0.1:3000/mcp. Set PORT to change
the port. It accepts modern and legacy MCP clients. It binds to loopback and validates
Host and browser Origin headers. There is no built-in authentication. For remote use,
put it behind your own HTTPS reverse proxy with access controls, preserve the backend
Host as 127.0.0.1:3000, and reject cross-origin browser requests at the proxy. Do not
forward arbitrary browser Origin values as trusted. The publisher provides no hosted URL.
Scope and evidence
The supported domain is currently credit risk. Market, operational, liquidity and other risk domains are not implemented. New tools can be added in later releases. This is an indicative single-exposure calculator, not a bank's approved model, regulatory reporting system, credit decision or investment recommendation.
IRB excludes PD=100%, financial-sector 1.25 multipliers, double default, collateral-specific CRR3 LGD floors for secured corporate and other retail exposures, foundation-approach collateral haircuts and portfolio LGD floors. Standardised calculations exclude credit risk mitigation, prior liens, SCRA sovereign floors, due-diligence adjustments and unimplemented national options. Bank-level output floors are not applied to individual exposures. Preferential conditions must be confirmed by the user.
Sources: EU CRR, CRR3 amendments, BDDK regulations. The formulas approximate the normal distribution, as documented on the website.
Verification is performed against independently calculated examples and through clean package installation with the official MCP client over stdio and HTTP. No bank-internal core, customer data or private repository files are distributed.
Methodology and limits · Privacy · Support
MIT licensed. Published by IRB Risk Weight Calculator.
Tool Schema Changelog
Recent tool additions, removals, and schema changes observed during successful MCP inspections.
Loading tool schema changelog…
This server cannot be deployed
Maintenance
Related MCP Connectors
A paid remote MCP for CLI tool MCP, built to return verdicts, receipts, usage logs, and audit-ready
Deterministic banking, LEI, VAT, SWIFT & compliance checks via MCP with signed XDR-1 receipts.
Self-hosted MCP server: 26 deterministic dev, security, and EVM tools.
Production MCP server for US equity and options intelligence: real-time IV radar, Monte Carlo simulation, options pressure, strategy backtesting, AI prediction, pre-trade risk analysis, and automated stock research reports.
Related MCP Servers
- FlicenseNot gradedqualityDmaintenanceMCP server that exposes the credit scoring model's deterministic tools (probability of default, SHAP explanations, typicality check, financial ratios) to AI agents, enabling natural language credit risk assessment.-
- FlicenseNot gradedqualityCmaintenanceLocal-first MCP server for credit risk analysis. It enables explainable rule discovery, statistical validation, PSI drift monitoring, and root-cause diagnosis on user-approved Parquet data while preserving privacy.-
- AlicenseBqualityCmaintenanceEnables AI agents to calculate Basel III SA-CCR exposure, run Monte Carlo PFE simulations, compute portfolio VaR/CVaR, and aggregate option Greeks through standard MCP tools and resources.4MIT
- AlicenseNot gradedqualityBmaintenanceEnables AI assistants to explain, scrub and check ISO 20022 payment messages and SSI files entirely on the local machine, including CBPR+ readiness hints, lifecycle triage, IBAN/BIC validation and payment-vs-SSI checks against the user's own licensed reference data. Runs as a local stdio MCP server with no network calls, telemetry or bundled vendor data.MIT