Clarity AI MCP Server

Official

Stress-test fund classifications against the proposed SFDR 2.0 rules from a chat prompt

Remote serverstreamable-http

What is the Clarity AI MCP server?

The Clarity AI MCP server is a narrow, purposeful connector: it exists to answer one hard question — how would this fund be categorised if the proposed SFDR 2.0 framework were in force today? Clarity AI is a sustainability data and analytics provider, and this hosted endpoint puts its regulatory modelling in front of an AI assistant so investment, product and compliance teams can run classification simulations conversationally instead of through a spreadsheet or a portal.

What it does

SFDR 2.0 is a proposed rewrite of the EU's sustainable finance disclosure regime, and the practical anxiety it creates is simple: funds currently marketed under one article may land in a different category under the new categorisation logic. This server lets you check. You give it a fund — by name, in plain language, or by an explicit ISIN — and it resolves the security, runs Clarity AI's SFDR 2.0 assessment against it, and returns an interpreted result rather than a wall of raw flags.

The tool design reflects real usage. There are two entry points depending on what you have: sfdr20_assess_fund_name is the one to use for free-text and partial names, because it resolves the fund first and then assesses it; sfdr20_check_by_isin is for when you already hold a precise identifier. The server's own tool descriptions are explicit that name-based requests should not be routed to the ISIN tool, which is a sign someone paid attention to how models mis-route in practice. A separate interpret_sfdr20_result tool normalises raw assessment flags into category-level evidence, so the answer comes with reasoning attached instead of a bare verdict.

Key capabilities

  • Fund resolution from natural language, partial names or exact identifiers such as ISIN and CUSIP
  • SFDR 2.0 category simulation returning an interpreted assessment rather than raw output
  • Evidence-level breakdown that shows which criteria drove a given category
  • Read-only tool surface — every tool is annotated read-only and non-destructive
  • Streamable HTTP transport with OAuth, so it drops into Claude, Claude Code, Cursor, Codex or VS Code without local installation

When to use it

This is a specialist instrument, not a general finance connector. It is genuinely useful if you are an asset manager mapping a fund range against the proposed categories, a compliance function preparing for reclassification work, or a distributor fielding questions about which products will still qualify. It is not a market data server, a portfolio analytics tool, or a way to pull ESG scores generally — asking it for anything outside SFDR 2.0 classification will not get you far.

That focus is a feature. Regulatory simulation is exactly the kind of task where a general-purpose model with web access produces confident nonsense, and where a vendor-maintained tool grounded in actual regulatory data earns its place in the loop.

Caveats worth knowing

Two tools, sfdr20_check_by_query and sfdr20_check, are deprecated aliases kept only for backward compatibility; new work should target sfdr20_assess_fund_name and sfdr20_check_by_isin instead. More importantly, the underlying tools describe themselves as hitting an upstream SFDR demo API, so treat the output as an indicative simulation of a framework that is still a proposal, not a compliance determination you can file against. Coverage is also bounded by Clarity AI's security universe — obscure or very new funds may not resolve at all. Finally, several tools accept an optional email parameter, which suggests results may be delivered or attributed per user, so read Clarity AI's terms before wiring this into an automated workflow.

What you can do with it

Map a fund range against the proposed categories

Walk a list of products through the assessment one by one and collect which would land in each proposed SFDR 2.0 category, so product teams can see where reclassification pressure sits before the rules land.

Explain a classification to a stakeholder

The interpretation tool returns category-level evidence rather than a bare label, which makes it straightforward to draft a note explaining why a fund assesses the way it does.

Resolve a fund you only half-remember

Free-text resolution means a partial name is enough to find the right security and its identifiers, which saves a trip to a separate reference data lookup.

Available tools

ToolWhat it does
search_securitiesSearch securities by keyword, ISIN or CUSIP against Clarity AI's SFDR data source.
sfdr20_assess_fund_namePreferred entry point: resolve a fund from free text or a partial name, run the SFDR 2.0 check and return an interpreted assessment.
sfdr20_check_by_isinRun the SFDR 2.0 check for a fund when you already have an explicit ISIN.
interpret_sfdr20_resultNormalise raw SFDR 2.0 flags into category-level assessment evidence.
sfdr20_check_by_queryDeprecated alias for sfdr20_assess_fund_name, retained for backward compatibility.
sfdr20_checkDeprecated alias for sfdr20_check_by_isin, retained for backward compatibility.

How to install the Clarity AI MCP server

claude mcp add clarity-ai --transport http https://clarity-sfdr20-mcp.pro.clarity.ai/mcp

Configuration

An MCP client that supports remote Streamable HTTP servers with OAuth. Access is subject to Clarity AI's terms for its SFDR 2.0 simulation service.

Example prompts to try

  • How would the Nordea 1 Global Climate and Environment Fund classify under SFDR 2.0, and what evidence drives that?
  • Look up the fund with ISIN LU0348926287 and run the SFDR 2.0 assessment.
  • Search for funds matching 'global sustainable equity' and show me their identifiers.
  • Compare the SFDR 2.0 assessment evidence for these three funds and summarise where they differ.

Frequently asked questions

It simulates how a fund would be categorised under the proposed SFDR 2.0 framework. You supply a fund by name or ISIN, the server resolves the security, applies Clarity AI's assessment logic and returns an interpreted result with category-level evidence rather than raw flags. It is scoped to that question — it is not a general ESG scoring or market data connector.