A unified database browser for 9 engines (SQLite, PostgreSQL, MySQL, MariaDB, SQL Server, MongoDB, Redis, Cassandra, Neo4j) exposed as MCP tools
picklock defines 6 tools with clear verb-noun naming (list_*, connect, browse, inspect, query). All tools have descriptions (avg ~120 chars, within baseline 34-392). Input schemas are present with type declarations. However, parameter descriptions are sparse or missing entirely, most params lack actionable guidance on format, constraints, or dependencies. Output schemas are undocumented. Error handling is minimal; no recovery guidance or actionable error messages visible. The 'query' tool accepts arbitrary statements with no validation hints, creating injection risk. Overall: solid naming and basic structure, but parameter and output documentation fall short of production grade.
Browse/paginate data from a database object (table rows, documents, key entries, graph nodes).
Connect to a database engine. Provide either a DSN string (e.g. postgres://user:pass@host:5432/db) or the name of a saved connection.
Inspect the structure/schema of a database object (columns, indexes, collection stats, key info, label properties).
List saved database connections from ~/.config/picklock/connections.json
List all objects in the connected database (tables, collections, keys, labels, depending on the engine).
Execute a query against the connected database. Supports SQL, MQL (MongoDB), RESP (Redis), CQL (Cassandra), and Cypher (Neo4j). Mutations are blocked in read-only mode.
Parameter descriptions missing or incomplete. 'dsn' param in connect lacks format guidance (which engines, example patterns). 'page' and 'page_size' in browse lack min/max constraints. 'statement' in query has no validation hints or injection warnings.
Output schemas undocumented. No specification of what fields list_connections, list_objects, browse, inspect, or query return. LLMs cannot plan downstream calls or extract required IDs without knowing response structure.
No error handling guidance. Tools do not document what errors are retryable, user-fixable, or fatal. 'query' tool with arbitrary statement execution has no injection prevention or error recovery hints.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | C | 63 | 2026-07-28+ | v2 |
Pagination not fully specified. 'browse' accepts page/page_size but no documentation of total count, next_cursor, or whether results are capped. Max 100 rows mentioned in code but not in tool description.
No input validation hints for 'query' tool. Arbitrary SQL/MQL/Cypher statements accepted with no guidance on injection prevention, read-only enforcement, or mutation blocking. LLMs cannot reason about safety.