IntelliJ IDE plugin that exposes database operations as MCP tools, enabling LLM-driven SQL query execution, schema inspection, and custom tool definitions against IDE-managed data sources.
IDE Database MCP has well-structured tool definitions with clear naming (verb_noun pattern) and comprehensive descriptions. All 8 tools have explicit schemas with typed parameters and descriptions. However, output schemas are not documented, error handling guidance is absent, and several parameter descriptions lack actionable constraints (e.g., enum values for 'scope', format specs for 'sql'). The server relies on caller discipline rather than validation, and destructive operations (DDL, NoSQL write) lack confirmation patterns. Descriptions are solid (100-300 chars) but could be more LLM-optimized with explicit prerequisites and recovery guidance.
Execute NoSQL statements intended for read/query usage. Tool selection is controlled by description and parameters only.
Execute NoSQL statements intended for write/delete usage. Tool selection is controlled by description and parameters only.
Execute SQL intended for schema/structure changes. Choose this tool for CREATE/ALTER/DROP style operations by convention.
Execute SQL intended for data modification. Choose this tool for INSERT/UPDATE/DELETE style operations by convention.
Execute SQL intended for read/query use and return rows. This tool relies on caller discipline via description and parameters; the server does not parse SQL keywords for enforcement.
List databases/catalogs/schemas under one IntelliJ-managed data source. When metadata is unavailable, fall back to query tools for manual inspection.
Output schemas not documented. Tools return rows/results but LLMs cannot infer field names, types, or structure. Downstream tool chaining is blind.
Destructive operations (DDL, NoSQL write) lack confirmation or dry-run patterns. Agents can execute DROP TABLE or DELETE without safeguards.
'scope' parameter accepts free-form strings (GLOBAL/PROJECT/ALL) but is not declared as an enum. LLMs may hallucinate invalid values like 'LOCAL' or 'SHARED'.
Inferred effective spec: <=2025-11-25.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | C | 63 | <=2025-11-25 | v2 |
Discover available IntelliJ Database data sources. Each entry includes name, url, driverClass and an inferred `type` label. Call this first when you need a valid dataSource name for other database tools.
Get the table and column structure (schema) of a database. Call this tool BEFORE writing SQL queries whenever you need to know what tables exist, what columns they have, or how they relate to each other. For large databases with many tables, provide keywords matching your current task (e.g. 'order', 'user', 'payment') — only the most relevant tables are returned, ranked by keyword matches in table names, column names, and comments, plus foreign-key connectivity. Without keywords the most interconnected (core) tables appear first. Use catalog/schema/tablePrefix to further narrow the scope. IMPORTANT: Use 'catalog' and/or 'schema' to specify the target database namespace. Omitting both causes a full scan across ALL available databases and may result in a timeout. Do NOT put the database/schema name into 'keywords'; keywords are for table/column name relevance scoring only.
SQL/NoSQL parameters lack format constraints or validation hints. Descriptions say 'intended for query/read' but do not explain what happens if LLM passes DML into query tool or vice versa.
No error handling guidance. Descriptions do not explain what to do if dataSource is invalid, SQL syntax fails, or connection times out. LLMs receive raw errors with no recovery path.