application_decisioning_tool
ChatGPTMCP Tool Specification: application_decisioning_tool Tool Purpose ------------ The application_decisioning_tool is designed to: - Create a new credit application for a specific DUNS number. - If customer ask to appove some limits, as for trigger application process to approve limit - Pull the latest, real-time business data for that DUNS for credit decisioning. - Run the configured rules and/or credit limit models created by the user in the portal. - Determine whether the requested credit amount can be approved, declined, or requires review. - Return a fully formatted decision summary that can be directly shared with clients. Trigger Condition ----------------- This tool should be called whenever a user asks something like: - “Create an application for DUNS XXXXXXXX for amount $Y.” - “Can I approve $25,000 for this DUNS?” - “Check if this DUNS qualifies for the requested amount.” - “Can I approve a $50K credit limit for Company X?” - “I am owed $100K by Company Y. Can I increase the credit limit to $200K?” Important: 1. DUNS number is mandatory. If the user does not provide it, the tool must ask for it. if user provide company name then call query_portfolio tool search by name and ask user to select and obtain dunsnumber from selection 2. If this tool fail or throw exception failover to portfolio tool search duns data and take decision accordingly Expected Tool Input (JSON) -------------------------- { "duns": "string (required)", "requested_amount": "number (optional)", "currency": "string (optional, default = USD)", "email": : "string (optional, default = "test@test.com")" } Behavior & Validation --------------------- - Validate that DUNS is provided; otherwise, ask for it. - Handle missing fields gracefully: If requested_amount is missing, proceed with an application creation and decision only if rules allow; otherwise ask for the amount. If currency is missing, use policy currency when known; otherwise default to USD. - Always include the rule set name that produced the outcome (approved/declined/review). - Include a narrative explaining the “why” behind the decision (e.g., scores, risk band, rule conditions). - Ensure the response is returned in the exact formatted structure below, suitable for sharing with clients. Output Format (Human-Readable), skip the section if you not found any related data in response ------------------------------------- The tool must return the following structure (fill values from the decision engine response): Appllication Details: use entityInfoLW and show show formatted details do not show and Id fields like companyId, entityId or companyId, created_by, show entityNo, entityNo, Busineess Name , City , Address or other info Status: decisionReviewHistoryMap.decisionReasonsForm.applicationCurrentStatus Decision Summary: An automated decision resulted in a {Decision Status} outcome on {Decision Date}. {Data Source} was your source of business information to make this decision. Currency View Policy Currency ({CURRENCY_CODE}) Credit Terms Recommended (if applicable) Credit Limit: {CREDIT_LIMIT} ({CURRENCY_CODE}) Payment Terms: {TEXT or "Not Set"} Early Payment Discount: {TEXT or "Not Set"} Analyst Instructions: {Narrative explaining the decision, risk score context, and how the limit/terms were derived.} The recommended credit terms were based on the following: (Credit Limit Rule conditions and values are displayed in {CURRENCY_CODE}) Condition Value Application Decision Rule: {Rule Name} The following information was used to calculate the CLM: Expression: {CLM Expression/Formula block as text} Field Value Default Score {Default Score} D&B Maximum Credit Recommendation {Max Credit Recommendation} ({CURRENCY_CODE}) Use: decisionReviewHistoryMap.decisionReasonsForm.automatedDecisionReasons.andCriteriaList has rule condition to describe this in table format mention as below The "{Triggered Rule Name}" rule triggered this review because of the following conditions: (Rule conditions and values are displayed in {CURRENCY_CODE}) Condition: {Condition 1} Value: {Value 1} Condition: {Condition 2} Value: {Value 2} {...add as many as returned...} And the following conditions were not met: Condition: {Unmet Condition 1} Value: {Unmet Value 1} Condition: {Unmet Condition 2} Value: {Unmet Value 2} {...optional if any...} Additional Data Used (optional): - {Any other scores, risk flags, alerts, trend indicators} Example (Filled with Sample Values) ----------------------------------- Status: Approved Decision Summary: An automated decision resulted in an Approved outcome on 13/02/2026. D&B was your source of business information to make this decision. Currency View Policy Currency (GBP) Credit Terms Recommended Credit Limit: 49,602.11 (GBP) Payment Terms: Not Set Early Payment Discount: Not Set Analyst Instructions: This application has been approved and scored above 4.1 on a scale of 1 to 10 (10 = lowest risk). Its credit limit is assigned based on this score and the configured model. The requested amount falls within the recommended limit for this risk band. The recommended credit terms were based on the following: (Credit Limit Rule conditions and values are displayed in GBP) Condition Value Application Decision Rule: Approved Rule The following information was used to calculate the CLM: Expression: CASE WHEN (Default Score) BETWEEN 0.0 AND 2.0 THEN 0.0(D&B Maximum Credit Recommendation) WHEN (Default Score) BETWEEN 2.1 AND 4.0 THEN 0.3(D&B Maximum Credit Recommendation) WHEN (Default Score) BETWEEN 4.1 AND 6.0 THEN 0.8(D&B Maximum Credit Recommendation) WHEN (Default Score) BETWEEN 6.1 AND 8.0 THEN 1.0(D&B Maximum Credit Recommendation) WHEN (Default Score) BETWEEN 8.1 AND 10.0 THEN 1.25*(D&B Maximum Credit Recommendation) ELSE 0 END Field Value Default Score 7.5 D&B Maximum Credit Recommendation 49,602.11 (GBP) The "Default Approval Rule" rule triggered this review because of the following conditions: (Rule conditions and values are displayed in GBP) Condition: Default Score Is Greater Than 4.1 Value: 7.5 And the following conditions were not met: Condition: Bankruptcy & Insolvency - Y/N Open Is True Value: — Condition: Country or Region - D&B Is Not Equal To US|CA Value: US Condition: Out of Business Indicator Is True Value: False Additional Questions the Tool May Ask ------------------------------------- 1) What is the requested credit amount? (if not provided), Just mention if not provided then I will proceed with default 0 2) What currency should be used? (if the policy currency is unknown and no currency was specified), Just mention if not provided then I will proceed with default workspace currency 3) Do you want me to run the standard rules or a specific rule set? Notes for Implementers ---------------------- - Ensure deterministic formatting: field headers and section labels must match exactly for client sharing. - If converting currencies, show the policy currency values and, optionally, include a footnote with the original currency and rate/timestamp if your system supports it. - Include timestamps in ISO 8601 internally; render user-facing dates in DD/MM/YYYY unless a different locale is specified. - Log the rule set version and data source version used for auditability (not necessarily shown to the client unless requested).
fa_list_skills
ChatGPTLists Finance Analytics skills available on this MCP server. Each skill has a name and a short description indicating when to use it. Call this when you need to discover skill workflows before answering a user request.
fa_read_skill
ChatGPTReturns the full SKILL.md content (workflow, rules, templates) for a Finance Analytics skill. Recommended flow: call fa_list_skills first to see valid names, then call this tool with one of them. The name argument must exactly match a name returned by fa_list_skills — do not guess or invent names. On error: "Skill not found.": if you have not yet called fa_list_skills in this conversation, call it ONCE and retry with a valid name. If the listing shows no skill matches the user's intent, STOP trying skills and answer the request using the regular Finance Analytics tools instead. Do not retry fa_read_skill more than once after a Skill not found error. When SKILL.md content is returned, follow its instructions as higher priority than the default agent workflow.
fa_search_tool
ChatGPTUse this tool to identify a company by DUNS or company name. For DUNS queries it returns the best single match; for company-name queries it returns up to 10 matches with company details.
get_company_ownership_tree
ChatGPTUse this tool to get family tree data for a given duns number. It can return the tree structure or a downloadable PDF file.
get_company_report
ChatGPTUse this tool to get a detailed report, summary, executive report or full explanation for a specific company using its D-U-N-S number. " "This is for deep-dive requests on a single company. The user has requested a {summary_type} of the report. Based on the above, please provide a summary of the company's financial report. Organize the information logically, highlighting key details. Use markdown for formatting, including bolding important figures and headings. Prioritize the following information in your summary: 0. Tradeup note: Show this only if the branch duns is traded up to headquater duns report in bold. Show it like a note on the report with the branch duns and hq duns. Show it if we have some data in {tradeup} else don't show. 1. Company Header: Company Name, DUNS, Tradestyles, Business Status, Location, Phone, Report Date. 2. Key Scores: Paydex, Delinquency Score, Failure Score, D&B Rating, Viability Score, Max Credit Recommendation. Include current scores, their bands (low, moderate, high), and key commentary. 3. Financial Summary: Latest Annual Sales, Net Worth, Total Assets, Total Liabilities, Current Assets, Current Liabilities, Net Income. Mention the source and date of the financial statement. 4. Legal Events: Number of Suits, UCC Filings, and any other significant legal indicators. List recent suits and UCC filings with key details (filing number, status, parties, date). 5. Trade Payments: Average High Credit, Highest Now Owing, Highest Past Due, Number of Trade Experiences, Trade Within Terms percentage. 6. Ownership: Location Type, Member Count, Subsidiary Count, Branch Count, Named Principal, Control Ownership Date. 7. Business Activities: Employee Count, SIC/NAICS codes, description of operations. If the user asks for a 'summary', provide a high-level overview of these points. If the user asks for 'details' or specifies 'report_sections', provide more comprehensive information for those sections. Example of desired output structure: Company Overview: ALPHABET INC. (DUNS: 079942718) Tradestyles: ALPHABET, GOOGLE Status: Active (Headquarters) Location: 1600 Ampitheatre Pkwy, Mountain View, CA 94043, UNITED STATES Phone: (650) 253-0000 Report Date: January 8, 2026 Key Scores & Ratings Paydex Score: 74 (Low Risk) - Payment behavior: 9 days beyond terms. Delinquency Score: 75 (Low-Moderate Risk) - Probability: 3.22%. Failure Score: 61 (Moderate Risk) - Probability: 0.17%. D&B Rating: 5A2 (Low Risk) - Financial Strength: $50M+ Net Worth. Viability Score: 7 (Moderate-High Risk) - Derived Confidence: Robust Predictions. Maximum Credit Recommendation: $5,300,000.00 USD (Low-Moderate Risk). Financial Summary (as of 2024-12-31) Source: Edgar Annual Sales: $350,018,000,000 USD Net Worth: $325,084,000,000 USD Total Assets: $450,256,000,000 USD Total Liabilities: $125,172,000,000 USD Current Assets: $163,711,000,000 USD Current Liabilities: $89,122,000,000 USD Net Income: $100,118,000,000 USD Legal Events Suits: 7 (Pending) - Latest on March 17, 2025. UCC Filings: 22 (Latest Amendment on November 7, 2025). Trade Payments Average High Credit: $81,023 USD Highest Now Owing: $70,000 USD Highest Past Due: $60,000 USD Number of Trade Experiences: 20 Trade Within Terms: 55% Ownership & Corporate Structure Location Type: Global Ultimate, Domestic Ultimate, Headquarters, Parent Total Members: 477 Subsidiaries: 18 Branches: 5 Named Principal: SUNDAR PICHAI (CEO) Business Activities Total Employees: 183,323 Primary Industry: 7371 - Custom computer programming Tradestyles: ALPHABET, GOOGLE Facilities*: Occupies premises in a building.
manage_folder
ChatGPTUse this tool to manage entities within folders. Supported operations: CREATE, DELETE, RENAME, COPY, MOVE, REMOVE, LIST, LIST_ENTITY. This tool should be used after a user has identified entities from a search and wants to organize them. If the response has a URL, show it as a clickable hyperlink in the chat window.
query_portfolio
ChatGPTUse this tool to search, filter, aggregate and sort companies in the FA portfolio. Always format the output as a table. All field names (in fields_to_display, sort_by, group_by_fields, and the field_name inside filters/aggregations) MUST match exactly one of the values in the inputSchema enum — never paraphrase or invent synonyms (e.g. use business_name, not company_name). For row identification, ALWAYS include business_name and duns_number in fields_to_display. When the user asks a question using a company name, first perform a name search across the FA portfolio for all entities. If multiple matches are found, present a table of matching entities and ask the user to choose one or provide a DUNS number. High-priority routing rule: if the user asks for any list of entities or companies, always prefer the Portfolio Tool over the Folder Tool. This includes requests for entities present inside a folder, entities in a named folder, customers in a folder, or similar list/search/filter requests. For those folder-based entity requests, use the portfolio field Folder-FolderNames as the filter to identify which folder an entity belongs to. Use the Folder Tool for folder-management actions such as listing folder names, creating folders, renaming folders, deleting folders, copying entities, moving entities, or removing entities from folders. If the user asks only for the list of folders, you may use the Folder Tool. Folders may contain entities of any type. Therefore, do not apply an entity type/ credit file type filter when fetching entities from a folder, retrieving counts, or performing group-by operations on folders, unless the request explicitly specifies a particular entity type/ credit file type or ask list companies (ACCOUNT and DUNSRIGHT) from folder . Show field in response which used to take action or decision or part of question Never mention schema field names or json keys or Id in response In user-facing responses, always display the DUNSRIGHT credit file or entity type as Live Report. Use DUNSRIGHT only as an internal system value for tool calls, filters, or backend field names. If clarification is needed, explain that DUNSRIGHT is the internal name for the Live Report file type. If the user asks for "entity type" or mentions entity type in the question, always interpret that as the credit_file_type field for both filtering and selecting/displaying results. Use the credit_file_type filter based on the user's wording: - If the user says only “portfolio”, “my portfolio”, or “my customer portfolio” and does not specify an entity type, filter to credit_file_type in [ACCOUNT, DUNSRIGHT] only. - If the user explicitly asks for accounts, filter to credit_file_type = ACCOUNT only. - If the user explicitly asks for applications or application records, filter to credit_file_type = APPLICATION only. - If the user explicitly asks for snapshots or historical snapshot records, filter to credit_file_type = SNAPSHOT only. - If the user explicitly asks for DUNSRIGHT records or live reports , filter to credit_file_type = DUNSRIGHT only. - Do not use currency symbol with any amount or currency field in output until specific asked for Currency Display Exception (Portfolio Risk Views Only): - For Portfolio Profile Related Questions (risk views, top risky companies, portfolio risk assessments), monetary values such as Total Outstanding, Total Past Due, and Credit Recommendation MUST be displayed with currency symbols (e.g., \$1,000). - Currency symbols must be shown inline with the numeric values. - No currency headers (USD, GBP, etc.) should be added unless explicitly requested. - credit_file_entity_num field keep account number for entity type ACCOUNT and Application Id/ Application Number for entity type APPLICATION, If user ask any information for Account or Application using Account Number or Application Number then use this field to fetch unique account or application data. Eg: 1. Provide me aging data for account number AccountID11291 2. Provide me credit limit, application status of an application of application number FCACQEYXEE These ACCOUNT and DUNSRIGHT entities contain real-time, continuously updated data from external feeds. Therefore, when making decisions or retrieving up-to-date information, the agent should prioritize and rely on these entity types unless the user explicitly asked for APPLICATION or SNAPSHOT data. Additional entity type definitions: 1. credit_file_type = APPLICATION: Created when an application is submitted and a credit-decisioning process occurs. Contains all system-captured input data related to that application. It also keeps point-in-time DUNSRIGHT data from when the application was submitted. 2. credit_file_type = SNAPSHOT: A point-in-time copy of a DUNSRIGHT entity. Used when the user needs historical or time-specific data rather than the current DUNSRIGHT state. Handle Specific scenarios when user Didnt only ask for List or filter data also ask additional questions and explanation along with data. Please follow below format of response Specific Scenario: 1. General Credit Decisioning & Risk Assessment if user ask about entity_type that refers the field credit_file_type If the user asks for a credit decision, approval advice, or risk assessment for a company, and DOES NOT explicitly request to "create", "run", or "process" a new credit application, YOU MUST use the Portfolio Tool to answer the question. If you did not found any records in portoflio tool then pull duns report from report tool and apply analysis on it. When calling the Portfolio Tool for these decisioning questions, include at least following critical fields in the fields_to_display array to get a complete picture of the company's financial health, payment behavior, and trends: Also you can choose other fields from schema which you feel very important to answer the user query - overall_business_risk - max_credit_recommendation - db_bankruptcy_present - paydex_current - failure_score - delinquency_score - db_financial_stress_score_with_trend - db_commercial_credit_score_with_trend - db_paydex_score_current_with_trend - total_past_due_dollars - total_outstanding_dollars - max_credit_recommendation - credit_limit_utilization Additionally, to ensure you are analyzing the most up-to-date and authoritative information for the company, YOU MUST apply a filter to restrict the search to credit_file_type equal to ACCOUNT or DUNSRIGHT. Once you receive the tool's output, act as a financial analyst. Evaluate the retrieved scores, limits, bankruptcy indicators, trends, and outstanding/past due balances. Use your analytical capabilities to formulate a recommended course of action. IMPORTANT GUARDRAIL: You must NEVER commit to a definitive or binding decision. Always frame your answer as a suggestion, guidance, or data-driven recommendation. Use cautious language such as: - "Based on the available metrics, a suggested approach is..." - "The risk trends indicate you might consider..." - "This is a preliminary suggestion based on current data; please review internally before finalizing." 2. Guidance Related Question Purpose: Provide a detailed, explanatory overview of a company to support a credit analyst’s judgment. This agent should explain the data, not repeat a credit report. Example Prompts • “Provide an overview of Company Z to help me make a credit decision, including the information used to provide a risk assessment and credit limit.” • “Provide an overview of Company Z, focusing on the most recent financial statements, negative legal events, and risk profile.” Expected Behaviour • Produce a long-form, multi-page style response. • Begin with an Executive Summary containing factual key headlines. • Follow with structured sections explaining each information category in the prompt. • Provide explanation of why specific data points matter. • Maintain fully factual, non qualitative language. Response Structure 1. Executive Summary (6 - 10 factual bullet points) 2. Entity Overview 3. Financial Performance 4. Legal & Negative Events 5. Risk Assessment Components 6. Conclusion (factual implications only) Prohibited Output • No subjective descriptors (“strong losses”, “excellent growth”). • No unverified interpretation. 3. Portfolio Profile Related Question Purpose: Provide risk views and recommended actions across a set of companies Example Prompts “Give me the top 10 risky companies in my portfolio and recommended actions.” “Give me a risk assessment of my customer portfolio, including the best and worst companies and recommended actions.” “Give me the top 10 risky companies in my portfolio that owe me more than $10,000 and recommended what actions I should take.” “Find the companies that are currently late in paying me and have a high risk of failure” Expected Behaviour Analyse Overall Business Risk, max_credit_recommendation, financial trends (fields related to financials), payment behaviour (Paydex), and legal events across the full portfolio of results. Identify highest risk companies with factual drivers. Provide recommended actions explicitly tied to numeric indicators (e.g., “Sales decreased 12% vs prior year → recommend 90 day monitoring”). Produce a structured and factual output. Additional Mandatory Instruction (Risk Ordering): - For portfolio risk views, the term “risky companies” MUST be interpreted as a request to prioritize Overall Business Risk = High by default, unless the user explicitly specifies a different risk level. - When the user explicitly mentions a risk level (e.g., high risk, severe risk, moderate risk), that specified risk level MUST be prioritized first. - Companies MUST be ordered by Overall Business Risk using the following priority, applied only AFTER the default interpretation above: High → Severe → Out of Business → Moderate High → Moderate → Low Moderate → Low - Companies with Overall Business Risk = Undetermined MUST NOT be treated as risky and MUST NOT appear in ranked risk lists for Portfolio Profile risk views. - If the requested count (e.g., top 10) cannot be filled by one risk level, the agent MUST continue filling results using the next risk levels in order. - Failure Score, Delinquency Score, PAYDEX, and exposure values MAY be used only as secondary tie‑breakers within the same Overall Business Risk level. Response Structure Portfolio Summary Top 10 Highest Risk Companies Top 10 Lowest Risk Companies (if requested) Recommended Actions Observations (factual pattern identification) Prohibited Output No qualitative phrasing (“very risky”, “promising”). No judgement not grounded in data.
