MCP App Store
by KeenEthics
Fitch Solutions icon

Fitch Solutions MCP Server for Claude and ChatGPT

By Fitch Solutions

Overview

Access trusted Fitch Ratings credit research and data, CreditSights intelligence, and BMI country and sector risk analysis — all through a single, governed MCP connection. Query ratings, financials, key rating drivers, research, and deal insights directly inside your AI workflows. Built for regulated environments with enterprise-grade authentication and entitlement controls. Fitch intelligence, where you work.

How to connect

  1. Claude: Open the Fitch Solutions listing in the Claude connectors directory and choose Connect. Claude listing Fitch Solutions exposes 18 tools here.
  2. ChatGPT: Open the Fitch Solutions listing in the ChatGPT app directory and choose Connect. ChatGPT listing Fitch Solutions exposes 34 tools here.

Tools

bmi-ask-research

ChatGPT
Answer questions about BMI published research, including country risk, macroeconomic and industry outlooks, market analysis, and sector reports. Use this tool when the user asks about BMI's views, analysis, or research on a specific country, market, sector, or topic. SCOPING TO ONE REPORT: if the user is asking about a specific report -- typically a follow-up to a get_research_list result, e.g. "summarise this report", "what are the key views in the Emerging Markets Monitor report", "what does report BMI_... say about inflation" -- pass that document's reportId as report_id. The answer is then drawn only from that report. If the user names a report by title but you do not have its id, call get_research_list with the title as topic first, then call this tool with the reportId it returns. Never guess or construct a report id. This tool is for QUALITATIVE questions -- views, outlook, drivers, "why", "what does BMI think about...", context and analysis. It is NOT the tool for a specific data point. If the user asks for a number, value, amount, count, rate, share, price, index/score or per-capita figure (e.g. "how many...", "how much...", "what is the total/average...", "what share of..."), use get_data_and_forecasts instead -- it returns the actual BMI time series. Reserve ask_research for the narrative around such figures. If the response has an 'error' field, report that error only — do not synthesize an answer from contexts. Otherwise, answer from the retrieved contexts. MANDATORY: Citation requirements When the tool result contains a non-empty citations array and you use one or more retrieved contexts to answer, end the answer with: Sources - [Title](url) (publishedDate) Example output format: [Answer text based on retrieved contexts.] Sources - [citation.title](citation.url) (citation.publishedDate) Rules: 1. Read the citations array from the tool result. 2. Include a source only when information from its retrieved context (the text or data returned by the tool) supports one or more claims in the answer. A report title alone is not enough. 3. Use the title as the markdown link text and citation.url as the link target. 4. Use citation.publishedDate in parentheses after each link. 5. Do not display raw URLs outside markdown links. 6. Do not create, guess, or modify URLs. 7. If citations is empty or no context was used, omit the Sources section. 8. If the tool result contains an error, return only that error and do not create a Sources section. 9. If you used any retrieved context, do not omit the Sources section.

bmi-get-data-and-forecasts

ChatGPT
Fetch BMI historical data and forecasts for one or more economic indicators, across one or more countries and/or regions. USE THIS TOOL (not ask_research) for any question asking for a NUMBER, VALUE, AMOUNT, COUNT, RATE, SHARE, PRICE, INDEX/SCORE or PER-CAPITA figure -- anything of the form "how many...", "how much...", "what is the total/average/size/level of...", "what share/percentage of...", "the trend in...", or a comparison of such figures between countries/years. These are data points, and this tool resolves them to a BMI field and returns the actual time series. Only route to ask_research when the user wants narrative analysis, views, drivers or outlook rather than a figure. If in doubt for a quantitative question, call THIS tool first. Fall back to ask_research only if it returns no usable series. Covers all 18,318 BMI fields across five categories: Macro-Economic (GDP, inflation, fiscal, debt, reserves, FX, labour), Industries (autos, banking, food & drink, medical devices, oil & gas, pharma, ports, telecoms, tourism, metals & mining and more), ESG Country (climate and disaster risk, governance, health, education), Energy Transition, and Politics. Field resolution distinguishes a level from its growth rate, a local-currency figure from a USD one, a total from a per-capita figure, and a 5-year from a 10-year CAGR. Pass 'transform' and 'currency' when the question implies them. If a field is genuinely ambiguous the tool does not stop to ask: it resolves the phrase to its top few candidate concepts and returns data for all of them. Each resolved field carries a 'hierarchy' string giving its full category path down to the field, e.g. "Macro-Economic / Risk Index / Political Risk Index". When one requested field expanded into several series (they share the same 'field' text and are marked 'ambiguous'), present ALL of them, each as its own table titled with its 'hierarchy' breadcrumb and rows of (year, value). Do not silently pick one and drop the rest, and do not fall back to ask_research -- show the candidates and let the user see the alternatives. Provide at least one country or region, unless every requested field is a globally-scoped commodity price/production figure (e.g. "gold price", "iron ore price") with no per-country breakdown at all -- Global is queried automatically for those, so omit location entirely for a pure commodity question rather than guessing a country. This also applies when a commodity field is mixed with a country-specific one in the same request (e.g. "gold price and Germany's unemployment rate") -- Global is added for the commodity field automatically alongside whatever location you pass for the other field. You do not need to add "Global" yourself. fields is optional: if omitted, the tool uses a default macro indicator set. Data is available on a rolling annual window of roughly the last 10 years through the next 9 years (recomputed relative to today). Use start_year and end_year to narrow the years returned. Requests fully outside the supported window are rejected with an explicit error. Only annual data is available, except the BIRI (Banking Industry Risk Indicator) fields -- BIRI Score, Financial Score, Government Finance Score, Living Standard Score, Regulatory Quality & Environment Score, International Linkages Score, Economic Volatility Score -- which are quarterly only and cover roughly the last 5 years (no forecasts). Each series is returned at its own frequency automatically, labelled by its 'frequency'. Quarterly points also carry a 'quarter' (1-4) and the quarter-end 'date' (e.g. "2025-09-30"), and have no forecastType. Label BIRI periods as quarters (e.g. "Q3 2025"), never as a year. If a question specifies a frequency, pass frequency so a mismatch (e.g. quarterly inflation) is rejected explicitly instead of silently returning another frequency. Reg…

bmi-get-projects

ChatGPT
Fetch project-level data from BMI's 5 project databases: Infrastructure Key Projects, Global Mines Projects, LNG Projects, Pharma Manufacturing Projects, and Data Centre Projects. USE THIS TOOL for questions about specific infrastructure, mining, LNG, pharma manufacturing, or data centre PROJECTS/FACILITIES -- their status, value, size, location, timeline, ownership, or the companies/contractors involved. This is project-level detail about individually named projects, not the macro/industry aggregate figures get_data_and_forecasts returns. Every attribute available for a matching project is returned. Up to 100 matching records are returned per call (merged across every database queried). 'total_count' always reports how many actually matched, even when it's more than 100, so a broad query is never silently incomplete -- it comes back as a labelled subset instead. If databases is omitted, every database you are entitled to is searched, and each result carries a 'database' tag so you can tell them apart. A database you are not entitled to is silently excluded from the search, not reported as an error -- only requesting an unrecognised database name is an error. If the user gives a specific project id (e.g. "lng_416", "mine_8803", "pharma_1", or a bare number with the database named separately), pass it as project_id instead of searching -- this fetches that one record directly and ignores countries/status/company/filters. An id with no recognised prefix needs a single value in databases to say which database it belongs to. IMPORTANT: If the tool response contains an 'error' key, report only that message and do not retry with different parameters. IMPORTANT: Present results as a table or list using the actual field values returned -- do not summarise away the data. Include the 'database' tag when results span more than one database so the source dataset is clear. IMPORTANT: If 'truncated' is true, tell the user how many projects matched in total ('total_count') and that you're showing the first 'returned_count' of them -- then suggest narrowing with an additional country/status/company/ filters value to see a more targeted set, rather than presenting the subset as if it were the complete answer.

bmi-get-research-list

ChatGPT
List BMI research documents that match the given criteria, with download links. Returns documents only -- titles, dates and links -- never an answer. Trigger phrases: "list BMI reports on...", "which BMI reports cover...", "show me BMI research about...", "find reports for [country/sector]...", "what has BMI published on...", "give me the latest BMI research for...", "download link for BMI reports on...", "browse BMI research on...". USE THIS TOOL when the user wants to find, browse or list research documents by topic, country, region or sector, so they can open the reports themselves. Do NOT use it when the user wants to know what BMI thinks, says, expects or forecasts -- "what is BMI's view on...", "what does BMI say about...", "why...", "summarise BMI's outlook for..." -- those are questions that need an ANSWER with citations, and belong to ask_research. If the user asks for a specific number, value or time series, use get_data_and_forecasts. At least one of topic, countries, regions or sectors must be provided. Countries and sectors are hard filters: only documents whose primary country / primary sector matches are returned, and always only within the user's content package. Topic and regions influence ranking only. MANDATORY OUTPUT RULES: - If the response has an 'error' field, report that error message only. - Otherwise present the 'results' array as a list of documents: one bullet per document, the title as a markdown link to its 'url', followed by 'publishedDate' in parentheses, and the country/sector when present. Never display raw URL strings. - Do NOT write a summary, synthesis, analysis or answer from the results. The response intentionally contains no document content. If the user then asks about one of the listed documents ("summarise this report", "what does it say about X"), call ask_research with that document's reportId as report_id so the answer is drawn from that report only. Returns: dict with 'count', 'results' [{title, url, reportId, publishedDate, country, sector}], 'filters_applied', 'query', and 'response_tool_time_ms'. On failure the dict has a single 'error' message instead.

cs-get-article-headlines

ChatGPT
Finds most recent 5 financial articles matching the given criteria. This tool is useful for the questions like: - Show me latest articles written by analyst John Doe - Give me recent new issue reports - Give me the list of recent {content_type_names} - What are the articles on {company_name} with {content_type_names} - How many articles are written by analyst John Doe - Give me the count of recent {content_type_names} :param: request containing criteria for finding articles :type: request FindArticlesRequest :return: One page of articles matching the given criteria :rtype: ArticlesPage

cs-get-articles-content

ChatGPT
Finds and returns relevant information from CreditSights (CS) articles that is most relevant to the given question. Articles are CreditSights' original written research and contain our analysts' perspectives on macro trends, credit strategy, relative value, new issues, issuer research, sector research, covenants analysis, recommendations and recommendation changes, and market news and developments. This is the primary tool for surfacing CreditSights' own analysis and opinions. USE THIS when the user asks about: CreditSights' view on an issuer, sector, market, or recent development, macro or strategy commentary, relative value across issuers or sectors, new issue analysis or new issue recommendations, sector research or sector outlooks, issuer specific research or credit analysis, covenant analysis or covenant commentary from our analysts, a recommendation change or rating action from CreditSights, recent market developments or news as covered by CreditSights, or any general research question where the user wants our take rather than a primary source or a transcript. DO NOT USE for questions about source documents like an offering memorandum, credit agreement, or lender presentation. Use get_source_documents for those. DO NOT USE for what management actually said on a lender call. Use get_transcripts for that. DO NOT USE for event scheduling, timing, or call access details. Use get_events for those. If a user asks a covenant question and wants our interpretation rather than original document language, use this tool first. FALLBACK: If this tool does not return a specific answer to a factual or numerical question (e.g. a figure, date, or disclosure from a company filing), call get_source_documents next before concluding the answer is unavailable. :param question: The user question in its original form if the question is self-contained. If the user question is not self-contained, feel free to rephrase the question and add information which will improve the search.

cs-get-business-profile

ChatGPT
Use this tool to retrieve the business profile for a company covered by CreditSights. The business profile covers company overview, business description, sector, and organizational structure. Call this tool when a user asks who a company is, what they do, or needs background context on a specific issuer."

cs-get-companies

ChatGPT
Finds companies/issuers for company profiles, research coverage, and risk analysis. Use this tool for queries about COMPANIES, ISSUERS, or BORROWERS: - "Show me companies in financial services" - "List technology companies" - "Find North American issuers" - "Companies with high yield rating" - "Borrowers in healthcare sector" - "Firms with strong credit risk scores" EXECUTION NOTE: Execute this tool directly without explanatory text. Return results immediately. Key patterns for ambiguous queries: - "Companies in [country]" → country_names: ['Algeria'] - "Companies in bonds" → asset_class_ids: ['bonds'] - "[Country] bond companies" → country_names: ['Algeria'] + asset_class_ids: ['bonds'] Note: "Companies", "issuers" search the same database. IMPORTANT: No tearsheet filtering available - only dockets_available (legal documents). DO NOT use this tool for: deal transactions, bond issuances, loan facilities, or financial instruments.

cs-get-core-scores

ChatGPT
Use this tool to retrieve proprietary CreditSights core scores for a company. Scores are quantitative ratings across key credit dimensions and can be returned as a score summary or chart. Call this tool when a user asks for a company's core score, credit scoring breakdown, or wants a scored assessment of an issuer."

cs-get-credit-views

ChatGPT
Returns the CreditSights Fundamental View on a company: our assessment of the company's financial fundamentals and credit profile specifically. Covers projected financial performance (revenue, EBITDA, margins, cash flow), leverage and credit metrics, and financial policy (debt targets, capital allocation, shareholder returns). Use this tool ONLY when a customer explicitly asks about a company's financials, fundamentals, credit metrics, leverage, or financial policy. Do NOT use this tool for general requests about our view, opinion, or recommendation on a company — those must return the CS View via get_cs_view. Despite the word "View" in its name, this tool does NOT contain our recommendation or overall stance on a company. Use when the customer asks things like: “what is Company X's fundamentals", “What is Company X's financial policy", “What is the fundamental view on Company X"

cs-get-cs-views

ChatGPT
Returns the CreditSights View (CS View) on a company. This is our analysts' recommendation and overall opinion on the company. Includes our recommendations such as Outperform, Marketperform, or Underperform. Also includes the qualitative thesis behind it such as strategic positioning, key catalysts and risks, management commentary, and our forward-looking take. This is CreditSights' PRIMARY opinion on a company and the DEFAULT response to any request for our view, opinion, recommendation, stance, take, or call. Use this tool whenever a customer asks for our view on a company, unless they explicitly ask for financial fundamentals or credit metrics (in which case use get_fundamental_view). Use when the customer asks things like: “what is your view on Company X", “what is CreditSights' view on X", "what's your call on Company X", “what is your opinion / recommendation / stance on Company X", "what do you think about Company X”.

cs-get-documents-list

ChatGPT
Search for and return a list of available Source Documents. This includes Lender Updates, Lender Presentations, Lender Decks, Offering Memorandums, Credit Agreements, Financial Statements, and other Deal Site documents. This tool provides an inventory of documents that exist for an issuer, deal, or document type. It returns document metadata such as title, type, publication date, issuer, and source link. It does not return content from within a document. This should be the default tool whenever the user wants to know what documents are available, rather than what a document says. USE THIS when the user asks about: - What documents are available for an issuer or deal - The latest or most recent documents on file - A list, inventory, or count of documents - Which lender updates, presentations, decks, memorandums, agreements, or financial statements exist for an issuer - Documents filed or published within a specific date range - Whether a specific type of document exists for an issuer ROUTING RULE If the user's question is asking whether documents exist, how many there are, or wants a list of them, use this tool first. Signal words that indicate this tool over get_source_documents: - give me - list - what documents - what do you have - latest documents - how many - do we have - any documents - show me all DO NOT USE when the user asks: - What are the highlights or key points from a document - What a document says about a specific provision, basket, or threshold - Financial figures found within a specific presentation or statement - To pull, summarize, or validate content from a document Use get_source_documents instead when the user wants information found inside a document, such as terms, figures, quotes, provisions, or summaries of what a document says. If the user hasn't named a specific company, ask them which one before calling this tool rather than guessing or omitting the issuer. Example prompts: - What documents do you have for Chromalloy? - List the lender presentations for Graham Packaging. - Show me all offering memorandums filed for Wesco since January 2025. - What credit agreements are on file for Acrisure? :param issuer: Company or issuer name to list documents for. Required. :param document_type: Optional filter, for example "Lender Presentation" or "Indenture". :param published_from: Optional ISO 8601 date (YYYY-MM-DD) lower bound. :param published_till: Optional ISO 8601 date (YYYY-MM-DD) upper bound. :param skip: Number of results to skip for pagination. :param take: Maximum number of results to return. Response formatting: present the returned documents as a markdown table with columns Title, Type, Date, Issuer. Always render Title as a markdown link to that document's source_url rather than plain, non-clickable text. If total_results exceeds the number of documents returned, say only the first N are shown and that narrowing document_type or the date range will surface more.

cs-get-event-risk

ChatGPT
Use this tool to retrieve the risk and catalyst assessment for a company covered by CreditSights. Covers event-driven risks and upside/downside catalysts that could affect the credit, such as M&A activity, refinancing risk, or covenant triggers. Call this tool when a user asks about event risks, catalysts, or credit triggers for a specific issuer.

cs-get-events

ChatGPT
Search CreditSights Events data for upcoming and historical events, including lender calls and commitment dates, along with full details such as date, time, company, event type, and description. Use this for any question about when a call is happening, how to join it, or what commitment dates are coming up. USE THIS when the user asks about: - Upcoming or historical events - Lender call schedules or commitment dates - What events are available for an issuer - When a lender call is taking place - Registration information, or how to join, register for, or dial into a call - What events are happening within a specific date range - Details for a specific event such as event name, event type, date, time, or description DO NOT USE for questions about what was discussed during a lender call after the call has occurred. Use get_transcripts for transcript content on past lender calls, speaker commentary, lender questions, or call discussion. DO NOT USE for CreditSights analysis, investment recommendations, or analyst views. Use the relevant CreditSights research or view tool for those questions. DO NOT USE for questions about a specific document, presentation, update, deck, filing, memorandum, or agreement, e.g.: - Lender updates or lender presentations - Lender decks - Offering memorandums - Credit agreements - Financial statements - Deal site documents Use get_source_documents for those. Routing rule: - If the question contains words like "lender update", "presentation", "deck", "financials", "offering memorandum", "credit agreement", "source document", or "deal site document", use get_source_documents instead. - If the user is asking what is contained in a document, use get_source_documents. - If the user can answer their question with a date, time, event description, registration link, or call access information, use get_events. - If the user is asking what was said or discussed on a call, use get_transcripts instead. Example prompts: - What lender calls are scheduled this week? - What is the next event for Acrisure? - Show me all upcoming commitment dates in August. - When is the next lender call for Life Time Fitness? - What are the details for the Acrisure lender call? - List all events for Latham Group over the next 30 days. - Are there any upcoming lender calls for healthcare issuers? - Show me the event description and call access information for the next lender call. :param company_names: Company names to filter events for (text search on company name). :param event_types: Filter by event type. Valid values: 'LenderCall', 'CommitmentDate'. :param event_start_date_from: ISO 8601 datetime string for start of date range (e.g. '2026-07-01T00:00:00.000Z'). :param event_start_date_to: ISO 8601 datetime string for end of date range (e.g. '2026-12-31T23:59:59.999Z'). :param title: Filter events by title or event name. Use whenever the user refers to a specific event by name or title. Can be used alone without company_names. :param sort_by: Sort field. Valid values: 'Title', 'EventStartDate', 'Company_Name', 'EventType'. :param sort_order: Sort direction. Valid values: 'Asc', 'Desc'. :param take: Maximum number of results to return.

cs-get-financials

ChatGPT
Search CreditSights Financials for structured company financial statement data, including Income Statement, Balance Sheet, Cash Flow, and KPI line items, broken out by fiscal year end (FYE) and fiscal quarter (Q) periods. This tool returns discrete, queryable financial figures (for example, revenue, gross profit, operating income, EBITDA, leverage ratios) rather than the underlying source document itself. Use this tool only when the user is asking for a specific financial figure, line item, metric, or trend that can be answered with a number, not when the user wants to read or summarize the original document a figure appears in. Priority Rule When user asks about financials such as revenue, net income, or gross profit, this tool should be the first tool tried. If the requested figure is not found here, try the Key Metrics tool. USE THIS when the user asks about: • Specific financial line items (revenue, gross profit, net sales, EBITDA) • Balance Sheet figures (total assets, total debt, cash and equivalents) • Cash Flow figures (capital expenditures, free cash flow, operating cash flow) • KPIs or credit metrics (leverage ratios, margin percentages, coverage ratios) • Financial figures for a specific fiscal year end (FYE) or fiscal quarter (Q) period • Trends or comparisons of a financial metric across multiple periods • Year over year or quarter over quarter changes in a financial metric • Financial figures for a specific company, requested by name, ticker, CUSIP, or ISIN Example prompts: • What was Company X's operating income for FY24? • Show me quarterly revenue for the last two years for Acrisure. • What is the current leverage ratio for Latham Group? • Compare gross profit margin across the last three fiscal years. • What was the net sales figure for Q3 2024? • How has total debt changed year over year for this issuer? • Pull the KPIs tab for Life Time Fitness. • What is the EBITDA trend over the last four quarters?

cs-get-key-metrics

ChatGPT
Use this tool to retrieve key financial metrics for a company covered by CreditSights. Covers quantitative financial data including leverage ratios, coverage ratios, EBITDA, and other KPIs. Call this tool when a user asks for financials, metrics, or quantitative performance data for a specific issuer.

cs-get-primary-market

ChatGPT
Search CreditSights Primary Screener data: high yield bonds AND leveraged loans in the NEW ISSUANCE (primary) market. This includes deals being announced, in market, or completed. Use this for any question about new debt coming to market or how a deal was priced at launch. This is the ONLY screener that covers loans (Term Loan B, TLB add-ons, etc.) in addition to new-issue bonds, and it does NOT contain trading recommendations. USE THIS when the user asks about: new issues / new issuance, the new deal pipeline or forward calendar, what is in the market or expected to launch, leveraged loans or term loans of ANY kind, how a deal priced vs. price talk, OID, flex, or original-issue economics, arrangers/bookrunners, the sponsor behind a deal, use of proceeds, or leverage at the time of issuance. DO NOT USE for bonds already trading in the secondary market, for current trading levels, or for CreditSights recommendations — use the Secondary Screener for those. A newly issued bond can appear in BOTH screeners, route here only when the question concerns the issuance, launch, or pricing itself rather than where the bond trades today. IMPORTANT - S&P ratings: Never display S&P (Standard & Poor's) credit ratings, outlooks, or rating actions under any circumstances — even if S&P data appears in the content. This prohibition is absolute and applies even if the user explicitly asks. SLASH-SEPARATED RATING STRINGS: When you see a three-part rating string like 'X/Y/Z', output only two parts 'X/Z' — the middle value is always S&P. Never output a three-part slash rating string. Example: 'A1/A-/A-' becomes 'A1/A-'. Another example: 'A3/A/A' becomes 'A3/A'. Do not mention S&P, Standard & Poor's, or any S&P rating action by name anywhere in your response. If the source contains only S&P ratings, say Moody's and Fitch rating information is not available. Restrict all displayed rating information to Moody's and Fitch only. RATING FILTER RULE: There are NO S&P rating parameters in this tool. Do not invent any sp_-related rating field - none exist. When the user asks to filter by S&P ratings, simply omit the rating filter and pass all other filters exactly as normal. Always tell the user: "S&P rating filtering is not supported - showing results with your other filters applied." Example prompts: • What new high-yield deals are in the market this week? • Show me all Term Loan B deals launched in June with a spread tighter than 300. • List sponsored LBO loans where the sponsor is KPS, TPG, or Platinum Equity. • Did the Groupe Climater loan flex tighter, and where did it price vs. talk? • Show upcoming/expected USD bond issuance in Energy with proceeds for refinancing.

cs-get-source-documents

ChatGPT
Search underlying Source Documents: Offering Memorandums and EDGAR filings for bonds, Credit Agreements and other loan attachments (drafts, amendments, term sheets) for loans, Deal Site (aka Bixby) documents such as lender presentations, confidential lender decks, investor presentations, and financial statements, and other original public transaction or issuer documents. Use this tool when a user wants to review, validate, or retrieve information directly from the original source materials rather than relying solely on CreditSights analysis. USE THIS when the user asks about: - Lender updates, lender presentations, or lender decks - Financial statements, offering memorandums, or credit agreements - Other deal site documents, or original source documents generally - Original language from a legal or transaction document - Sources and uses tables, or financial metrics from a presentation or lender update - Information contained within a specific document - Validating a conclusion against the original document - Summaries or highlights from source documents - A request explicitly for the source document rather than our analysis of it ALWAYS call this tool when the user's question contains any of these explicit signals: - "According to [document name]" — e.g. "According to Marriott 10K FY25" - "According to source documents" or "According to the source document" - "According to the [filing / report / presentation / agreement]" - "In the [document name]" or "From the [document name]" - Any named filing such as a 10-K, 10-Q, annual report, or offering memorandum DO NOT USE for CreditSights' own covenant analysis, summary, or interpretation of a deal. Use the Covenant Review report for those by default (get_articles_content / get_article_headlines). If a user asks a general covenant question without asking for original language, reference the Covenant Review report first, not this tool. DO NOT USE for questions about when a lender call is scheduled, how to join a lender call, or event and commitment dates. Use get_events for those. DO NOT USE for questions about what management said during a lender call, quotes or questions from a lender call, or transcript excerpts. Use get_transcripts for those. DO NOT USE for questions asking what documents are available for a company, such as "what documents do you have for X" or "list source documents for X", when the user isn't asking about the content of a specific document. Use get_documents_list for those. Routing rule: - If the question contains words like "lender update", "lender presentation", "lender deck", "source document", "offering memorandum", "OM", "credit agreement", "financial statement", "investor presentation", "management presentation", "deal site document", "quarterly update", "monthly update", or "board presentation", use this tool first — as long as the information requested is from a document rather than a lender call transcript. - If the user is asking when a call is scheduled or how to join it, use get_events. - If the user is asking what was said or discussed on a call, use get_transcripts. Combined queries: when a user asks a question that draws on both CreditSights' analysis and the underlying document (for example, comparing our summary against the original language, or validating a figure from a lender presentation against CreditSights' reported number), call both get_source_documents and get_articles_content and combine the results in a single response. Citations: each result's metadata identifies which source it came from along with the document's title and URL, so it's always clear when an answer is backed by a source document rather than CreditSights' own analysis. At the end of your response, always include a "Sources:" section listing the document title and URL for every res…

cs-get-transcripts

ChatGPT
Search CreditSights Transcripts data for lender call transcripts. These are full text records of lender calls that have already taken place, broken out by speaker and timestamp, along with metadata such as issuer, transcript name, and deal site. Use this for any question about what management actually said on a call. USE THIS when the user asks about: - Lender calls, or management commentary, guidance, or outlook delivered verbally on a call - What was discussed or which topics came up on a lender call - Questions and answers discussed on a call - Quotes or statements attributed to a specific speaker - Tone or sentiment across one or more calls - A summary of a prior earnings or lender call - Comparisons across multiple lender call transcripts DO NOT USE for questions about when a lender call is scheduled or what events are upcoming. Use get_events for future events, event timing, event discovery, and event details. DO NOT USE for CreditSights analysis, recommendations, or analyst views unless the user explicitly asks to compare transcript commentary with CreditSights research. DO NOT USE for questions about a specific document, presentation, update, deck, filing, memorandum, or agreement, e.g.: - Lender updates or lender presentations - Lender decks - Offering memorandums - Credit agreements - Financial statements - Deal site documents Use get_source_documents for those. DO NOT USE for questions asking what transcripts are available for a company, such as "what transcripts do you have for X" or "list transcripts for X", when the user isn't asking about the content of a specific call. Use get_transcripts_list for those. Routing rule: - If the question contains words like "lender update", "presentation", "deck", "financials", "offering memorandum", "credit agreement", "source document", or "deal site document", use get_source_documents instead. - If the user is asking what is contained in a document, use get_source_documents. - If the user can answer their question with a date, time, registration link, call access information, or other event detail, use get_events instead. - If the user is asking what was said or discussed on a call, use get_transcripts. Example prompts: - What did management say about liquidity during the latest Acrisure lender call? - Summarize the key takeaways from the First Advantage Q2 lender call. - What questions did lenders ask about leverage? - Quote management's comments on margin expansion. - What did the CFO say about refinancing plans? - Compare management commentary from the last two lender calls. - Show me all transcript excerpts where management discussed customer demand. - How has management's outlook changed across recent lender calls? :param question: The user question in its original form if the question is self-contained. If the user question is not self-contained, feel free to rephrase the question and add information which will improve the search.

cs-get-transcripts-list

ChatGPT
Search for and return a list of available transcripts. This tool provides an inventory of transcripts that exist for a date range, issuer, or deal site. It returns transcript metadata such as transcript name, deal site and source link. It does not return content from within a transcript. This should be the default tool whenever the user wants to know what transcripts are available, rather than what was said on a call. USE THIS when the user asks about: - What transcripts are available for an issuer or deal - The latest or most recent transcripts on file - A list, inventory, or count of transcripts - Transcripts filed or published within a specific date range - Whether a transcript exists for a specific call or issuer ROUTING RULE If the user's question is asking whether transcripts exist, how many there are, or wants a list of them, use this tool first. DO NOT USE when the user asks: - What management said during a lender call - Quotes from a lender call - Questions asked on a lender call - Summaries or highlights of what was discussed on a call - Specific financial figures or guidance mentioned during a call Use get_transcripts instead when the user wants information found inside a transcript, such as quotes, statements, discussion points, or answers to questions asked during a call. If the user hasn't named a specific company, ask them which one before calling this tool rather than guessing or omitting the issuer. Example prompts: - Give me the latest transcripts for Chromalloy. - Do we have a transcript for the Q2 lender call for Graham Packaging? - List all lender call transcripts filed for Wesco in the last 12 months. - How many transcripts have been published for this issuer this year? - Show me all transcripts available for this deal. :param issuer: Company or issuer name to list transcripts for. Required. :param call_type: Optional filter, for example "Lender Call" or "Bank Meeting". :param published_from: Optional ISO 8601 date (YYYY-MM-DD) lower bound. :param published_till: Optional ISO 8601 date (YYYY-MM-DD) upper bound. :param skip: Number of results to skip for pagination. :param take: Maximum number of results to return. Response formatting: present the returned transcripts as a markdown table with columns Title, Call Type, Date, Issuer, Deal Site. Always render Title as a markdown link to that transcript's source_url rather than plain, non-clickable text - it opens the transcript on the deal site screener. If total_results exceeds the number of transcripts returned, say only the first N are shown and that narrowing call_type or the date range will surface more.

rdr-ask-research

ChatGPT
Answer questions about Fitch published research, including sector and economic outlooks, issuer analysis, summary, rating methodologies, and industry reports. Use this tool when the user asks about Fitch's views, analysis, or research on a specific company, sector, or topic. Also use this tool, not get_research_list, when the question asks what a Fitch report says or asks for a figure, table value, or loan or property detail published inside one, including for a named deal, transaction, or trust. get_research_list returns only titles and download links and cannot read a document's contents. If get_research_list was already called (this turn or an earlier one) and resolved the specific document(s) the user means, pass their source_id via doc_ids to scope the answer to exactly those documents instead of a broad search. Trigger keywords: Fitch view, Fitch research, Fitch analysis, Fitch opinion, Fitch commentary, Fitch outlook, Fitch sector view, Fitch methodology, Fitch special report, Fitch thesis, Fitch take, Fitch assessment, contents of the presale report, within the report, loan or property details, rating case loss. Trigger phrases: "what does Fitch think about...", "what is Fitch's view on...", "what is Fitch saying about...", "Fitch's outlook for the [sector] sector", "summarize Fitch's research on...", "how does Fitch analyze...", "Fitch's take on...", "provide information on the [loan or property] in the [deal] transaction using the [report]", "give me the contents of the [report]", "provide information from the [table] within the [report]", "what does the [report] say about...". Also use for a bare "research", "analysis", "commentary", or "outlook" request when no provider is named. If the user names another provider, defer to that provider's tool. When citing sources use markdown link format only: Source Title. Never display raw URL's. Never show full URL string in the response text. If the response includes a 'citations' array, render each citation as a markdown link using its title and url: [title](url). Do NOT fabricate links for documents not listed in citations. Args: question: A specific, self-contained question (e.g. "What is outlook on the US banking sector?" or "How does Fitch rate sovereign debt?") doc_ids: Optional source_id values to scope the answer to specific documents. Returns: dict with context, sources and metadata which can be used to answer the question

rdr-get-entities

ChatGPT
Screen and discover Fitch-rated entities filtered by market sector, geography, and/or rating level. Use this tool when the user asks to screen, discover, or list entities by sector, region, or rating. At least sector_names and geographies (regions) must be provided. Trigger keywords: Fitch-rated entities, Fitch-rated companies, Fitch-rated banks, Fitch universe, Fitch coverage, screen ratings, screen, list, discover, filter, sector, region, geography, rating band. Trigger phrases: "list Fitch-rated [banks/corporates] in [region]", "which [sector] entities does Fitch rate", "screen Fitch ratings by [rating/sector/region]", "who does Fitch rate in...", "find Fitch-rated issuers rated [X] or above", "show Fitch coverage for [sector/region]". This tool screens a universe by filter criteria. For a single named issuer, use get_ratings, get_financial, or get_krd instead. This tool returns entity names and their current ratings only. To fulfil requests for additional data, use the returned entity names with: - get_financial: for key credit metrics (leverage, EBITDA, CET1, etc.) - get_krd: for positive and negative rating sensitivities - get_ratings: for full rating history and outlook Present only the 10 most relevant entities to the user based on best match to their query intent. Args: sector_names: Market sector names resolved server-side to sector IDs (parent + sub-sectors) geographies: ISO country codes (expand regions to individual codes) rating_codes: Fitch rating scale codes to filter by

rdr-get-financial

ChatGPT
Retrieve Fitch financial data for one or more issuers across any Fitch-covered sector (corporates, banks, sovereigns, insurance) — Fitch financials, metrics, statements, Fitch-adjusted figures, and ratios. Use this tool when the user asks about an entity's Fitch financial metrics, statements, or ratios, or any quantitative financial query about an issuer. Trigger keywords: Fitch financials, Fitch financial data, Fitch-adjusted, fundamentals, financial performance, income statement, balance sheet, revenue, EBITDA, leverage ratio, net income, margin, CET1, Tier 1, coverage ratio, debt, premiums, cash flow. Trigger phrases: "give me [entity]'s Fitch financials", "Fitch-adjusted EBITDA or leverage for...", "what is [entity]'s EBITDA for fiscal year...", "show me the leverage ratio for...", "what is the debt figure on [entity]'s balance sheet". Also use for a bare "fundamentals", "financials", or "financial metrics" request when no provider is named. If the user names CreditSights, defer to that provider's tool. Use the ratings tool for credit rating values, or the research assistant for broader qualitative analysis. Sector detection is handled automatically from the resolved entity. Entity names and financial concept terms must be extracted from the user's question by the caller and passed via entity_list and financial_fields respectively. Args: entity_list: Required list of issuer names taken exactly as the user wrote them in the query. Do not expand or modify them. Example: ["Apple Inc", "Microsoft Corp"] years: Optional list of years to retrieve data for, in YYYY format. Defaults to the last 6 years computed from today's date server-side. OMIT this parameter when the user asks for 'last N years', 'latest', or 'recent' — never hardcode a year range for open-ended recency requests, as that will produce stale results. When presenting 'last N years' results, ALWAYS show the N most recent (newest) years available in the dataset, newest first. Do NOT show the first N entries or an arbitrary year range. Years need not be consecutive, count from the newest year backward. Example: if data exists [2020, 2021, 2023, 2024, 2025] and user asks 'last 3 years', show [2025, 2024, 2023] (the 3 newest), NOT [2021, 2022, 2023]. Always include all years in the selected range, including Standardized-only years marked with adjMissing. For a follow-up about the same result, preserve the original entity, financial metric, and requested number of years unless the user changes them. Only pass explicit years when the user specifies exact years. Example: [2021, 2022, 2023] financial_fields: Optional list of financial concept terms extracted from the question. If omitted, a broad default set of fields is returned. Example: ["revenue", "ebitda", "net_income"] Returns: List of dictionaries with entity name, agent_id, and financial data. Corporate sectors — each metric is returned as ONE merged field per metric, with both value variants aligned by year. Each value entry includes: - adjustedValue: the Fitch-Adjusted value, or null when unavailable - standardizedValue: the Standardized value, or null when unavailable - standardizedValues: all Standardized values when a period has duplicates - adjMissing: true (only present if ADJ was unavailable for that year) - sameValueAsStd: true when adjustedValue and standardizedValue are equal IMPORTANT — default corporate presentation (when the user mentions neither variant): MUST use a table with the columns "Year, Fitch-Adjusted, Standardized", in that order. Read those columns from adjustedValue and standardizedValue respectively. Show null as —. Present the requested number of newest years first and do NOT omit years that have only Standardized data. When adjMissing is true, note that Fitch-Adjusted was unavailable. After the table, include a brief "Key observations" section descr…

rdr-get-issues

ChatGPT
Retrieve Fitch-rated issue-level data for bonds, notes, and other debt instruments. Use this tool when the user asks about specific debt issues, bonds, or notes for one or more issuers, or wants to filter by instrument type, debt level, CUSIP, or ISIN. Trigger keywords: Fitch-rated bond, Fitch-rated note, issue rating, instrument rating, bond, note, debt instrument, senior secured, senior unsecured, subordinated, coupon, maturity, CUSIP, ISIN, tranche, issuance. Trigger phrases: "list Fitch-rated bonds for...", "what bonds of [entity] does Fitch rate", "Fitch issue rating for [CUSIP/ISIN]", "[entity]'s senior secured notes rated by Fitch", "Fitch-rated debt instruments for...", "show [entity]'s bonds and their Fitch ratings". Use get_ratings instead for issuer-level ratings rather than instrument-level ratings. At least one filter must be provided. entity_names are resolved automatically to issuerIds. Results are grouped by issuer, with each issue containing only: id, classTypeDescription, bondName, debtLevelDescription, coupon, currency, isin, cusipWithPlacement, isinWithPlacement, and maturityDate. Args: entity_names: List of issuer names (e.g. ["Amazon", "Home Depot"]). issue_id: Fitch issue ID. class_type_description: Instrument class, e.g. "bond/note". debt_level_description: Debt seniority, e.g. "senior secured". cusip: CUSIP code, e.g. "931142AU7". isin: ISIN code, e.g. "AU3FN0040309". Returns: dict with issuers array (each containing issuer_name, issuer_id, issue_count, issues), or issue_count and issues array for non-issuer filters.

rdr-get-krd

ChatGPT
Retrieve Fitch Key Rating Drivers (KRD) — the factors that support or constrain an issuer's Fitch rating, including Fitch rating sensitivities. KRDs are Fitch-published summaries explaining the key factors that support or constrain an issuer's credit rating. Use this tool when the user wants to understand why an entity has its current rating, what factors could lead to a rating change, or any question about rating rationale, key rating drivers, rating sensitivities, or upgrade/downgrade triggers. Trigger keywords: Fitch key rating drivers, KRD, rating rationale, rating sensitivities, rating factors, positive factors, negative factors, downgrade trigger, upgrade trigger, credit opinion. Trigger phrases: "why is [company] rated...", "what are the key rating drivers for...", "what are Fitch's rating sensitivities for...", "what would trigger an upgrade/downgrade for...", "what factors support the rating for...". Also use for a bare "credit opinion", "rationale", "rating drivers", or "why is it rated X" when no provider is named. If the user names another provider, defer to that provider's tool. When citing sources use markdown link format only: Source Title. Never display raw URL's. Never show full URL string in the response text. If the response includes a 'citations' array, render each citation as a markdown link using its title and url: [title](url). Do NOT fabricate links for documents not listed in citations. Use the ratings tool instead if the user only wants the rating value or history. Use the research assistant tool for broader sector or thematic questions. Entity names — partial or informal names are accepted but canonical legal names yield the most reliable results. Args: entity_list: List of issuer names to retrieve KRDs for (e.g. ["Goldman Sachs", "Toyota"]). Maximum 10 entities. Full legal entity names are preferred over tickers or abbreviations. limit: Maximum number of KRD documents to return per entity (default: 1 for latest document). Increase to retrieve older documents for historical trend analysis. offset: Number of documents to skip (default: 0) Returns: Dictionary with documents list, count, and optional error. If multiple entities resolved, returns data for all entities.

rdr-get-peers

ChatGPT
Retrieve Fitch peer entities and comparables for one or more issuers. Use this tool when the user asks which companies are considered peers or comparables to a given issuer, or wants to benchmark an entity against similar firms. Trigger keywords: Fitch peers, Fitch peer group, Fitch comparables, comparable companies, peer set, peer group, peers, comparables, benchmark peers, Fitch-defined peers, similar issuers. Trigger phrases: "who are [entity]'s Fitch peers", "Fitch peer group for...", "comparable companies to [entity] per Fitch", "benchmark [entity] against Fitch-rated peers", "who does Fitch consider comparable to...", "peers of [entity] according to Fitch". Also use for a bare "who are [entity]'s peers" or "comparables for [entity]" when no provider is named. If the user names another provider, defer to that provider's tool. Use the financials tool to compare actual metrics across peers once retrieved. Use the ratings tool to compare credit ratings across peers. Use get_entities to screen a sector or region rather than peers of a named issuer. Sector is auto-detected from each entity's profile where possible. The sector argument acts as a fallback only — providing the correct sector improves accuracy when auto-detection is ambiguous. Args: entity_list: List of issuer names to retrieve peers for (e.g. ["JPMorgan Chase", "HSBC"]). Maximum 10 entities. Full legal entity names are preferred over tickers or abbreviations. sector: Fallback sector if auto-detection fails. Must be one of: 'banks', 'corps', 'sovereigns', 'insurance' (default: 'corps'). Note: each entity's actual sector is auto-detected from its profile and will override this value where possible. Returns: Dictionary with entity_count, sector, and results array containing peers for each entity. Example: {"entity_count": 1, "sector": "banks", "results": [{"entity_id": "123", "entity_name": "JPMorgan Chase", "peers": [...]}]}

rdr-get-ratings

ChatGPT
Retrieve Fitch credit ratings and rating history for one or more issuers — current or historical Fitch rating, IDR, outlook, or watch status. Fitch is an NRSRO. For a bare "credit rating", "the rating", "rating and outlook", or "rating agency view" when no provider is named, use this tool. Trigger keywords: Fitch rating, Fitch credit rating, Fitch IDR, Issuer Default Rating, Fitch long-term rating, Fitch short-term rating, Fitch outlook, rating outlook, watch status, rating watch, Fitch grade, Fitch-rated, current rating, rating history, rating change, downgrade, upgrade, notch, AAA, AA, BBB, CCC. Trigger phrases: "what is Fitch's rating on...", "how is [entity] rated by Fitch", "current Fitch rating for...", "Fitch rating history of...", "is [entity] on Fitch watch", "what's [entity]'s Fitch outlook", "did Fitch upgrade/downgrade...", "rated by Fitch", "was [company] downgraded", "show me the rating history for...". Use this tool for any question about what a rating is or was. Any question about the value, history, or status of a credit rating should route here. Questions about why a rating is what it is should route to get_krd instead. Args: entity_list: List of issuer names to look up (e.g. ["Apple Inc", "Ford Motor Co"]). Maximum 10 entities per request. Names should be explicit legal entity names, not abbreviations or ticker symbols. start_date: Start of the rating history window. Format: "YYYY-MM-DD". Defaults to "2016-01-01". Cannot be earlier than 2016-01-01. end_date: End of the rating history window. Format: "YYYY-MM-DD". Optional. Must be within 10 years of start_date. Returns: dict: Ratings history per entity, including long-term issuer rating, outlook (Positive/Stable/Negative), watch status, and effective dates. If multiple entities are resolved, results are returned for all of them.

rdr-get-research-list

ChatGPT
Search and list Fitch Ratings research documents with direct download links. Trigger keywords: list Fitch reports, browse Fitch research, Fitch publications, download link, recent research, research note, publication list. Trigger phrases: "list Fitch reports on…", "recent Fitch research for [entity/sector]", "find Fitch rating action commentaries on…", "download Fitch's report on…", "show me Fitch publications about…", "what Fitch reports exist for [entity]", "pull the latest research note on…", "get the latest reports for [entity] to download…", "find publications on…". Use this tool when the user wants to find or browse a LIST of research reports by criteria such as issuer, series, category, country, market sector, or publication date (e.g. "Show me recent reports for Apple", "Find ESG research from last month", "List rating action commentaries on US banks"). For a synthesized answer drawn from research content — i.e. the user wants to know what a report says, concludes, or found, or wants the model to reason over, extract data from, or analyze a SPECIFIC named report or document — use the ask_research tool instead, even if the request also references a topic, issuer, series, or category that this tool could filter on. This includes requests that anchor on one named report and ask the model to: pull specific figures, values, or table contents out of it, assess or explain risks, concentrations, or scenarios described within it, or judge whether the report's narrative is consistent with its own data, or with the issuer's broader history or strategy, or rank, summarize, or extract a list of items (e.g. transactions, metrics, risk factors) from a table or section inside it. The common signal is a request scoped to the contents of one identified document, phrased as "based on/using/in/from/within [report]...", "the contents of [report]", or "what does [report] say..." — as opposed to a request to browse or list reports matching a topic, issuer, or filter. Use get_research_list only when the user wants a list of reports to browse themselves (titles, dates, download links), not analysis or extraction from within a specific report's contents. Do not repeatedly loop on this tool if you cannot find the exact research report the user wants. Instead just return the list of research results which was found. If ask_research needs to answer about one of the reports found here, pass its source_id via ask_research's doc_ids argument to scope the answer to that document. At least one filter (topic, entity_list, series, category, countries, sector_names, published_to or published_from) should be provided to narrow results. Source is always Fitch Ratings and results are sorted newest-first by publication date. When the user asks for reports ABOUT a subject, event, or theme (e.g. "give me reports on Iran War"), set the 'topic' argument to that subject ("Iran War"). When 'topic' is set, a title search is run in addition to the standard filter query and the two result sets are merged (deduped by source_id), and the other filters below still apply to the standard query. Choosing series (preferred way to route a question): - Ratings, rating actions, or an issuer-specific outlook (outlook on a named issuer/entity, e.g. "outlook on Amazon") then series "Rating Action Commentary". - Issuer-specific research (about a named issuer/entity) then series "Issuer Research". - Ratings criteria / methodology questions then series "Ratings Criteria". - Any other general topic then series "Non-Rating Action Commentary". Avoid setting category unless the user specifically asks for a rating report ("Rating Report"), an upgrade ("Upgrade"), a downgrade ("Downgrade"), or a sector/industry outlook ("Outlook", e.g. "outlook on technology", "outlook…

rdr-sf-compare-cmbs-deals

ChatGPT
Compare CMBS deal performance — trend analysis and WA metric aggregation across deals. Compares issuance vs current metrics across CMBS transactions to identify deteriorating deals, compute sector averages, and rank transactions by any WA deal metric. All deal fields are returned, enriched with delta columns and flags when trend analysis is requested. For field definitions, see get-cmbs-deals. For deal identifiers, use fitch_deal_id or fitch_deal_name. OBJECT JOINS: use fitch_deal_id from results with get-cmbs-loans to drill into individual loan performance. Use this tool when users ask about: - Which CMBS deals have experienced the largest increase in cap rates and value decline since issuance? → flag_cap_rate_expansion=true, sort_by_metric=true, metric_field='fitch_wa_cap_rate_current' - Show me CMBS deals where debt yields have deteriorated (declined) since issuance → flag_dy_decline=true - Which deals have LTV increasing since issuance (pool deterioration signal)? → flag_ltv_increase=true - Identifying deals with worsening qualitative risk scores (QRS score going up = more risk) → flag_qrs_deterioration=true - Comparing CRE CLO deals from 2025 vintage by issuance Fitch WA LTV → market_sector='CRE CLO', review_year=2025, metric_field='fitch_wa_ltv_issuance' - What is the average WA cap rate across CMBS deals reviewed in 2024? → metric_field='fitch_wa_cap_rate_issuance', review_year=2024 - Compare average debt yield by market sector across all deals → metric_field='fitch_wa_dy_issuance', group_by_sector=true - Ranking CMBS transactions by current BSF rating-case loss → metric_field='deal_bsf_rating_case_loss_current_pct', sort_by_metric=true - Show average FLOC percentage across all CRE CLO deals → market_sector='CRE CLO', metric_field='floc_current_pct' - Compare CRE CLO vs CMBS deals on average issuance LTV → metric_field='fitch_wa_ltv_issuance', group_by_sector=true - Which deals show simultaneous cap rate expansion AND debt yield decline (dual stress signal)? → flag_cap_rate_expansion=true, flag_dy_decline=true When trend flags are set, each deal is enriched with a "trend_analysis" object containing: - cap_rate_delta, cap_rate_expanded (if flag_cap_rate_expansion) - ltv_delta, ltv_increased (if flag_ltv_increase) - dy_delta, dy_declined (if flag_dy_decline) - qrs_delta, qrs_worsened (if flag_qrs_deterioration) - any_trend_worsened: true if any metric worsened Returns: { "deals": [deal objects enriched with trend deltas and flags], "total_count": number of deals matched, "total_elements": total deals across all pages, "metric_field": the aggregated metric (if requested), "overall_aggregation": {avg, min, max, count} for metric_field, "breakdown_by_sector": per-sector aggregation (if group_by_sector=true), "filters_applied": active filter parameters, "response_tool_time_ms": execution time in ms }

rdr-sf-compare-cmbs-loans

ChatGPT
Compare CMBS loan performance — issuance vs current trend analysis AND metric aggregation. Supports two modes (combinable): 1. Trend flags: flag loans where a metric worsened since issuance (current vs issuance comparison) 2. Metric aggregation: compute avg/min/max/count for any numeric loan field across a segment All loan fields are returned, enriched with a "trend_analysis" object when any flag is set. For field definitions, see get-cmbs-loans. For loan identifiers, use fitch_loan_id or fitch_loan_name. OBJECT JOINS: use fitch_deal_id from results with get-cmbs-deals, or fitch_loan_id with get-cmbs-properties. Use this tool when users ask about: TREND / ISSUANCE vs CURRENT COMPARISONS: - Which loans show the largest LTV deterioration since issuance? → flag_ltv_increase=true - Find loans with widening cap rates since issuance (value pressure signal) → flag_cap_rate_expansion=true - Show loans where DSCR has declined since issuance (debt service stress) → flag_dscr_decline=true - Identify loans where NCF has deteriorated since issuance (operating income decline) → flag_ncf_decline=true - Find loans with declining property values since issuance → flag_value_decline=true - Show loans where BSF loss expectations have increased since issuance → flag_bsf_loss_increase=true - Identify loans with worsening qualitative risk scores (QRS going up = more risk) → flag_qrs_deterioration=true - Which hotel loans analyzed in 2025 show simultaneous cap rate expansion and DSCR decline? → property_type='Hotel', analysis_year=2025, flag_cap_rate_expansion=true, flag_dscr_decline=true METRIC AGGREGATION: - What is the average Fitch NCF haircut at issuance for multifamily properties in 2025 CRE CLO? → metric_field='fitch_ncf_haircut_issuance', property_type='Multifamily', market_sector='CRE CLO', analysis_year=2025 - What is the average issuance Fitch cap rate for hotel properties reviewed in 2025? → metric_field='fitch_cap_rate_issuance', property_type='Hotel', analysis_year=2025 - What is the average LTV for office loans at issuance? → metric_field='fitch_ltv_issuance', property_type='Office' - Compare average BSF rating-case loss by property type → metric_field='fitch_bsf_rating_case_loss_current_pct', group_by_property_type=true - Average NCF haircut by property type across 2025 vintage loans → metric_field='fitch_ncf_haircut_issuance', group_by_property_type=true, analysis_year=2025 - Showing average DSCR for multifamily loans reviewed in 2024 → metric_field='total_term_dscr_issuance', property_type='Multifamily', analysis_year=2024 When trend flags are set, each loan is enriched with a "trend_analysis" object containing: - ltv_delta, ltv_increased (if flag_ltv_increase) - cap_rate_delta, cap_rate_expanded (if flag_cap_rate_expansion) - dy_delta, dy_declined (if flag_dy_decline) - dscr_delta, dscr_declined (if flag_dscr_decline) - ncf_delta, ncf_declined (if flag_ncf_decline) - value_delta, value_declined (if flag_value_decline) - bsf_loss_delta, bsf_loss_increased (if flag_bsf_loss_increase) - qrs_delta, qrs_worsened (if flag_qrs_deterioration) - any_trend_worsened: true if any metric worsened Loans with worsened metrics are sorted first. Returns: { "loans": [loan objects enriched with trend_analysis when flags set], "total_loans_matched": total loans matched, "metric_field": the aggregated metric (if requested), "overall_aggregation": {avg, min, max, count} (if metric_field set), "breakdown_by_property_type": per-type aggregation (if group_by_property_type=true), "total_elements": total loans across all pages, "filters_applied": active filter parameters, "response_tool_time_ms": execution time in ms }

rdr-sf-get-cmbs-deals

ChatGPT
CMBS deal-level analytical surveillance and performance tool. Retrieves comprehensive Fitch deal metrics, risk indicators, and optional narrative commentary for Commercial Mortgage-Backed Securities transactions. Supports screening, surveillance, and deal profiling workflows. PRIMARY IDENTIFIERS (use these to look up a deal): fitch_deal_id — Unique Fitch integer ID for the deal (e.g. 96880527). Use deal_ids parameter. fitch_deal_name — Fitch-designated CMBS transaction name (e.g. 'CSAIL 2016-C7'). Use deal_name parameter. identifiers — List of CUSIP/ISIN identifiers for tranches in the deal. Use identifiers parameter. FIELD DEFINITIONS (all fields returned in each deal record): fitch_deal_id — Unique Fitch identifier for the Deal fitch_deal_name — Fitch designated CMBS transaction name fitch_analysis_date — Date of Fitch's most recent review, typically coinciding with the Last Rating Action date last_rating_action_comment_date — Date of Fitch's most recent Rating Action Comment for any tranche in the deal deal_bsf_rating_case_loss_issuance_pct — Fitch 'Bsf' rating-case loss at issuance. Includes pool concentration add-ons for multiborrower/agency deals deal_bsf_rating_case_loss_current_pct — Fitch 'Bsf' rating-case loss as reflected in current ratings at Last Rating Action Date floc_current_pct — % of current pool balance classified as Fitch Loans of Concern (FLOC). Loans with DSCR/occupancy/tenancy issues or on servicer watch lists pct_of_pool_credit_opinion_issuance — % of pool balance (at issuance) with investment-grade credit characteristics on a standalone basis pct_of_pool_credit_opinion_current — % of pool balance (at Last Rating Action) with investment-grade credit characteristics on a standalone basis fitch_value_decline_issuance — Weighted average difference between Fitch's trust value and the banker's appraised value at issuance cash_flow_variance_issuance — Weighted average difference between Fitch's trust NCF and the issuer's NCF at issuance fitch_wa_dy_issuance — Weighted average Fitch Debt Yield at issuance (Fitch NCF / cumulative proceeds or trust+pari passu balance) fitch_wa_dy_current — Weighted average Fitch Debt Yield at most recent rating action fitch_wa_constant_issuance — Weighted average Fitch-assigned loan constants at issuance fitch_wa_constant_current — Weighted average Fitch-assigned loan constants at current analysis date fitch_wa_cap_rate_issuance — Weighted average Fitch loan cap rates at issuance fitch_wa_cap_rate_current — Weighted average Fitch loan cap rates at current analysis date fitch_wa_ltv_issuance — Weighted average Fitch loan-to-value ratio at issuance fitch_wa_ltv_current — Weighted average Fitch loan-to-value ratio at most recent review wa_vr_score_issuance — Weighted average Value Resiliency (VR) score at issuance, measuring difference between Fitch cap rate and benchmark VR cap rate wa_vr_score_current — Weighted average Value Resiliency (VR) score at most recent analysis date wa_qrs_issuance — Weighted average Qualitative Risk Score (QRS) at issuance (scale 1-5: 1=least risky, 5=most risky) wa_qrs_current — Weighted average Qualitative Risk Score (QRS) at most recent analysis date (scale 1-5) wa_lss_issuance — Weighted average Loan Structure Score (LSS) at issuance (scale 1-3: 1=credit, 3=penalty) wa_lss_current — Weighted average Loan Structure Score (LSS) at most recent analysis date (scale 1-3) region — Geographic region of the transaction (e.g. 'North America', 'Europe') marketSector — Transaction type / market sector (e.g. 'Multiborrower', 'Single Borrower', 'CRE CLO', 'Agency') identifiers — List of CUSIP/ISIN tranche identifiers OBJECT JOINS (how Deal connects to other objects): Deal → Loans : use get-cmbs-loans with deal_name or deal_ids to get all loans in a deal Deal → Properties: use get-cmbs-properties with loan_ids from the loans to get underlying properties Loan → Deal : each loan record contains fitch_deal_id linking back to its deal Use thi…

rdr-sf-get-cmbs-loans

ChatGPT
CMBS loan-level analytical surveillance and performance tool. Retrieves comprehensive Fitch loan metrics, FLOC screening, NCF haircut analysis, and optional narrative commentary for individual loans within CMBS transactions. PRIMARY IDENTIFIERS (use these to look up a loan): fitch_loan_id — Unique Fitch integer ID for the loan. Use loan_ids parameter. fitch_loan_name — Name of the loan in the CMBS transaction. Supports case-insensitive, partial/substring matches. fitch_deal_id — Unique Fitch identifier for the parent deal. Use deal_ids parameter. master_servicer_loan_id — Master servicer's unique ID assigned to the loan (Trepp-generated). FIELD DEFINITIONS (all fields returned in each loan record): fitch_property_type — Property type per Fitch criteria (e.g. 'Multifamily', 'Office', 'Office-Urban', 'Retail', 'Hotel', 'Industrial') fitch_property_grade_issuance — Fitch on-site inspection quality grade at issuance (scale A-D: A=best quality, B=neutral/uninspected, D=lowest) fitch_analysis_date — Date of Fitch's most recent review, coinciding with Last Rating Action date fitch_ltv_issuance / fitch_ltv_current — Fitch LTV: loan principal / Fitch property value, at issuance and most recent analysis total_current_ltv_issuance/current — Fitch LTV based on current loan balances at issuance and most recent analysis total_maturity_ltv_issuance/current — Fitch LTV based on Fitch Maturity Balance at issuance and most recent analysis trust_maturity_ltv_issuance/current — Fitch trust LTV calculated at maturity trust_midpoint_ltv_issuance/current — Fitch LTV at assumed default midpoint (between current date and maturity) total_term_dscr_issuance/current — Term debt service coverage ratio at issuance and most recent review total_maturity_dscr_issuance/current — Maturity DSCR at issuance and most recent review fitch_cap_rate_issuance/current — Fitch-assigned capitalization rate for the property at issuance and most recent analysis fitch_market_cap_rate_issuance/current — Fitch's market cap rate assumption, blended with Fitch cap rate in certain leverage metrics fitch_ncf_issuance/current — Fitch Net Cash Flow (NCF) at issuance and most recent analysis fitch_ncf_haircut_issuance/current — Overall effect of Fitch adjustments to property NCF relative to historical or issuer cash flows fitch_haircut_driver_issuance — Primary driver of NCF haircut (e.g. 'vacancy', 'real estate taxes', 'rent decline', 'management fee') fitch_dy_issuance/current — Fitch Debt Yield: Fitch NCF / cumulative proceeds (single borrower) or trust+pari passu balance (multiborrower) fitch_value_issuance/current — Fitch-determined property value = Fitch NCF / Fitch cap rate fitch_value_decline_issuance — Difference between Fitch trust value and banker-appraised value at issuance fitch_cutoff_balance_issuance — Outstanding loan balance at the securitization cutoff date fitch_maturity_balance — Projected loan balance at maturity based on current analysis fitch_bsf_rating_case_loss_issuance_pct — Loan-level Bsf rating-case expected loss at issuance (before pool concentration add-ons) fitch_bsf_rating_case_loss_current_pct — Loan-level Bsf rating-case expected loss at most recent analysis adj_pd_lgd_issuance/current — Adjustments to PD or LGD not captured by standard Bsf rating-case assumptions term_pd_bsf_case_issuance/current_pct — Probability of default during loan term at Bsf rating case term_lgd_bsf_case_issuance/current_pct — Loss given default if loan defaults during term at Bsf rating case maturity_pd_bsf_case_issuance/current_pct — Probability of default at loan maturity at Bsf rating case maturity_lgd_bsf_case_issuance/current_pct — Loss given default if loan defaults at maturity at Bsf rating case credit_opinion_issuance/current — Boolean: whether loan has investment-grade credit characteristics on standalone basis qrs_issuance/current — Qualitative Risk Score (scale 1-5: 1=least risky, 5=most risky). Accounts for borrower strength, market conditions,…

rdr-sf-get-cmbs-properties

ChatGPT
CMBS property-level details, geographic concentration, size analysis, and deal diversification tool. Retrieves collateral property data for CMBS loans and supports geographic analysis, size ranking, and deal-level geographic diversification assessment. PRIMARY IDENTIFIERS (use these to look up a property): fitch_property_id — Unique Fitch integer ID for the property. Use property_ids parameter. fitch_loan_id — Unique Fitch identifier for the parent loan. Use loan_ids parameter. gwc_deal_id — Unique Fitch identifier for the parent deal. FIELD DEFINITIONS (all fields returned in each property record): fitch_property_type — Fitch-assigned property type (e.g. 'Multifamily', 'Office', 'Office-Urban', 'Retail', 'Hotel', 'Industrial') fitch_property_city — City where the property is located fitch_property_state — US state where the property is located (e.g. 'NY', 'CA', 'TX') fitch_property_msa — Metropolitan Statistical Area (MSA) designation for the property's location fitch_property_size — Total square footage or number of units at issuance region — Geographic region of the transaction (e.g. 'North America', 'Europe') marketSector — Transaction type / market sector (e.g. 'Multiborrower', 'CRE CLO') identifiers — List of CUSIP/ISIN tranche identifiers for the deal asOfDate — As-of date for the property record (yyyy-MM-dd) OBJECT JOINS (how Property connects to other objects): Property → Loan : use fitch_loan_id with get-cmbs-loans to retrieve loan performance metrics Property → Deal : use gwc_deal_id with get-cmbs-deals to retrieve deal-level metrics Loan → Properties: pass loan_ids to this tool to get all collateral for a given loan Use this tool when users ask about: - Which CMBS deals have the greatest exposure to retail (or any) property type? → property_type='Retail', deal_diversification=true, diversification_sort_by='property_count' - Identifying large CMBS properties in New York City → city='New York', size_order='desc' - Geographic breakdown of CMBS properties by state → geographic_group_by='state' - Identifying CMBS properties concentrated in major metro areas (top MSAs) → geographic_group_by='msa', top_n=10 - Finding properties with the largest footprint grouped by state or property type → size_group_by='state' (or 'property_type') - Identifying geographically diversified CMBS deals (spread across many states) → deal_diversification=true, min_states=5, diversification_sort_by='diversification' - Finding single-state or highly concentrated CMBS deals (geographic concentration risk) → deal_diversification=true, max_states=1 - Ranking deals by number of distinct states in their collateral pool → deal_diversification=true, diversification_sort_by='diversification' - Show CMBS retail properties across Northeast US states → property_type='Retail', geographic_group_by='state' - Property-level details for a specific loan or deal → loan_ids=['37740'] or property_ids=['143809'] Response shape depends on active mode: - Default (no grouping flags): flat list of property objects with all fields - geographic_group_by set: ranked concentration table (property_count, total_size_sqft, unique_loan_count per group) - size_order set: properties sorted by fitch_property_size - size_group_by set: avg/min/max size statistics per group (state/city/msa/property_type) - deal_diversification=true: per-deal summary (distinct_state_count, distinct_city_count, property_count, states list) Returns: { "properties": [flat list of property objects] (default mode), "concentration": [ranked geographic table] (geographic_group_by mode), "size_distribution": [grouped size stats] (size_group_by mode), "deals": [deal diversification summary] (deal_diversification mode), "total_count": number of records returned, "total_elements&quo…

rdr-sf-resolve-cmbs-fields

ChatGPT
Resolve business language or an allowed field ID before calling a CMBS data tool.

cs-get-article-headlines

Claude

cs-get-articles-content

Claude

cs-get-business-profile

Claude

cs-get-companies

Claude

cs-get-core-scores

Claude

cs-get-credit-views

Claude

cs-get-cs-views

Claude

cs-get-deals-and-tranches

Claude

cs-get-event-risk

Claude

cs-get-key-metrics

Claude

rdr-ask-research

Claude

rdr-get-entities

Claude

rdr-get-financial

Claude

rdr-get-issues

Claude

rdr-get-krd

Claude

rdr-get-peers

Claude

rdr-get-ratings

Claude

rdr-get-research-list

Claude

Similar Apps

Listed on MCP App Store

Listed on MCP App Store

Add this badge to your site or README. It links to this page.

HTML

<a href="https://mcp-app-store.keenethics-labs.com/app/fitch-solutions"><img src="https://mcp-app-store.keenethics-labs.com/badges/listed-on-mcp-app-store.svg" alt="Listed on MCP App Store" width="220" height="44"></a>

Markdown

[![Listed on MCP App Store](https://mcp-app-store.keenethics-labs.com/badges/listed-on-mcp-app-store.svg)](https://mcp-app-store.keenethics-labs.com/app/fitch-solutions)