Investment Banking MCP server for IPO lifecycle management, providing deep-tier business intelligence, data lineage audits, and multi-factor strategic checks across a medallion architecture (Bronze/Silver/Gold data tiers).
Critical deficiencies across all dimensions. Tool schemas are empty objects with no properties defined, LLMs cannot infer parameter intent. Descriptions are generic and lack actionable context (e.g., 'Fetches end-to-end data lineage' does not explain WHEN to call it or WHAT it returns). Most tools (7/9) have zero parameter documentation. No output schemas documented. Error handling is absent, tool calls return raw JSON with no recovery guidance. Naming is inconsistent: 'perform_self_healing_audit' conflates detection and correction (should split); 'get_strategic_valuation_matrix' is vague on what 'matrix' means. No tool annotations (readOnlyHint, destructiveHint). This server would fail code review.
Fetches end-to-end data lineage and integrity metadata.
Checks deal status and legal logs for IPO filing readiness.
Monitors deal distribution across Managing Directors and Associates.
Performs multi-factor scoring to determine deal risk and complexity.
Generates strategic IPO charts.
Fetches business-engineered deal profiles from the Gold tier.
Detects and autonomously corrects data anomalies in the fabric.
Empty input schemas: 7 of 9 tools define inputSchema as bare {'type':'object'} with no properties. LLMs cannot determine what parameters to pass or what they control.
No parameter descriptions: Tools with parameters (search_kb, generate_oracle_report) lack descriptions for 'query' and 'data_type'. LLMs cannot infer valid values or constraints.
Generic, non-actionable descriptions: Most tool descriptions (e.g., 'Fetches business-engineered deal profiles') lack WHEN to use, WHAT it returns, or dependencies. Descriptions average ~40 chars; baseline is 194 chars.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | F | 32 | 2026-07-28+ | v2 |
Searches the banking policy knowledge base.
Analyzes revenue vs. market valuation targets for deal feasibility.
No output schemas documented: Tool responses return raw JSON (e.g., 'content': [{'type': 'text', 'text': json.dumps(...)}]). LLMs cannot plan downstream calls or extract structured fields.
No error handling or recovery guidance: Tool calls return isError=true with bare text. No actionable next steps, no categorization (retryable vs fatal), no invalid value details.