HTTP-based MCP server for d-migrate database schema migration and reverse engineering tool. Provides tools for schema reversal, generation, comparison, and artifact management across multiple database dialects (PostgreSQL, MSSQL, MySQL, SQLite, Oracle).
Critical gaps across all dimensions. Tool definitions are inferred from shell scripts (mcp-client.sh, roundtrip-smoke.sh) rather than visible in actual MCP server code. No explicit tool registration, schema definitions, or descriptions found in source. Schemas are partially reconstructed from curl calls in bash, making them unreliable. Descriptions exist but are generic (10-50 chars, well below 50-200 char baseline for LLM optimization). Parameters lack type constraints and validation guidance. No error handling patterns, no output schema documentation, no security considerations visible. This is a shell-script wrapper around an undocumented HTTP endpoint, not a properly structured MCP server.
Retrieve a chunk of artifact content
List server capabilities and metadata
Get the status of a running or completed job
Compare two schemas and report differences
Generate DDL for a schema in a target database dialect
List schemas in the state store with pagination support
Start a schema reverse engineering job for a database connection
Tool definitions inferred from shell scripts, not visible in actual MCP server implementation. No explicit tool registration with schemas found in source code.
Input schemas incomplete or missing. job_status_get and artifact_chunk_get have no visible schema definitions. schema_list, schema_reverse_start, schema_generate lack type constraints (e.g., no enum for targetDialect values, no min/max for pageSize).
Descriptions are generic and under 50 characters (baseline 50-200 for LLM optimization). 'List server capabilities and metadata' (34 chars), 'Get the status of a running or completed job' (45 chars) lack context for when/why to call.
Inferred effective spec: <=2025-11-25.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | F | 31 | <=2025-11-25 | v2 |
No output schema documentation. Tools return JSON but structure is undocumented. LLMs cannot plan downstream calls or extract required fields (e.g., what does schema_list return? Is it {schemas: [...]} or {items: [...]}?).
No error handling guidance. Shell scripts use generic curl error handling; no recovery hints for LLMs. E.g., if schema_reverse_start fails, should agent retry, call a different tool, or ask user?