> ## Documentation Index
> Fetch the complete documentation index at: https://docs.creditbenchmark.com/llms.txt
> Use this file to discover all available pages before exploring further.

# Credit Benchmark MCP Server

> Bring live Credit Benchmark consensus credit data into Claude, Microsoft 365 Copilot, ChatGPT, and other AI assistants through the hosted Credit Benchmark MCP server.

The Credit Benchmark MCP server connects AI assistants such as Claude, Microsoft 365 Copilot, and ChatGPT directly to Credit Benchmark data. Ask about a company, a portfolio, or a sector in plain language, and the assistant looks up the Credit Consensus Rating, probability of default, and trend from Credit Benchmark before it answers.

It uses the [Model Context Protocol](https://modelcontextprotocol.io) (MCP), an open standard for connecting AI assistants to external data. The server is hosted by Credit Benchmark, so there is nothing to download or install.

> **[Connect your assistant](/mcp/connecting)**
>
> Add the server to Claude, Microsoft 365 Copilot, ChatGPT, Claude Code, Cursor, or VS Code.

> **[Tools](/mcp/tools)**
>
> The tools and prompts the server gives your assistant.

## Request access

The MCP server is in alpha, and access is not enabled automatically for Credit Benchmark accounts. To request access, email [support@creditbenchmark.com](mailto:support@creditbenchmark.com) with the names and email addresses of the users who need it. Our support team enables each account and confirms when it is ready.

Once your account is enabled, follow [Connecting](/mcp/connecting) to add the server to your AI assistant.

## What you can ask

* "What's the credit picture for Tesco?"
* "Write a credit memo exhibit for Volkswagen AG, including how our rating compares with consensus."
* "Here are 40 counterparties. Which ones have been downgraded in the last six months?"
* "Show the rating distribution of European banks, and how it has shifted over two years."
* "How has the aggregate PD for US utilities moved since last year?"
* "How is the Credit Consensus Rating calculated?"

The assistant chooses which tools to call. For a single company it typically resolves the name to a `CB_ID`, then shows a credit memo exhibit. For a list of names it resolves them in batches and then summarises the whole portfolio, rather than pulling every name one by one.

## How it works

**1. You ask a question**

You ask your assistant about a company, portfolio, or sector.

**2. The assistant calls the MCP server**

The assistant picks the relevant Credit Benchmark tools, such as name matching, entity data, or portfolio analytics, and calls them at `https://mcp.creditbenchmark.com/mcp`.

**3. The server queries Credit Benchmark**

The server calls the Credit Benchmark API on your behalf, using the account you signed in with.

**4. The assistant answers**

The assistant answers from the data returned. Single-name questions can display an interactive credit memo exhibit, and large results come back as a downloadable CSV.

## Access and data

* **Sign-in:** you connect with your Credit Benchmark account through Okta. The assistant holds a token that refreshes automatically, and never sees your password.
* **Entitlements:** results follow your existing Credit Benchmark entitlements. You see the same coverage, metrics, and history as in the Web App and API.
* **Primary-source data:** answers are based on anonymised consensus data contributed by around 40 global financial institutions, not on web search or news.

See the [MCP Privacy Policy](/mcp/privacy-policy) for what the MCP server collects, how long it is kept, and who it is shared with.


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.