Loading ToolBench
The dlt MCP server provides 10 tools with reasonable names and descriptions, but several tools lack parameter descriptions and detailed output schemas. Tools follow a consistent verb_noun pattern (list_, get_, execute_, display_, search_). Most tools are READ_ONLY and well-scoped. However, parameter documentation is inconsistent, some tools like execute_sql_query and search_docs have enumerated values and format guidance, while others like get_table_schema_diff have optional parameters without clear usage patterns. Output schemas are not documented in the source, making it difficult for LLMs to know what fields to expect. Error handling is not visible in the provided code. The prompts feature is implemented, showing additional agent-guidance capability beyond basic tools.
Generate a mermaid diagram to represent the pipeline schema
Executes SELECT SQL statement for simple data analysis. Use the `list_tables()` and `get_table_schema()` tools to retrieve the available tables and columns.
Retrieve metadata about data loaded with dlt.
Retrieve the pipeline state information. Includes: incremental dates, resource state, source state
Get the schema of the specified table.
Get the diff between schema versions of a table.
List all available dlt pipelines. Each pipeline has several tables.
Output schemas not documented. Tools return data but the structure is not specified in tool definitions, forcing LLMs to guess at response fields and chain parameters.
Parameter descriptions missing or vague. 'another_version_hash' in get_table_schema_diff lacks guidance on how to obtain a schema version hash; 'hide_columns' boolean in display_schema lacks explanation of when it should be used.
No error handling guidance visible. No documentation of what errors can occur, whether they are retryable, or what recovery steps the agent should take.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-22 | D | 58 | 2026-07-28+ | v2 |
| 2026-03-09 | F | 0 | - | v1 |
List all available tables in the specified pipeline.
Search over the `dlt` source code.
Search over the `dlt` documentation. Use it to verify if a feature exists, answer general questions, or identify recommended patterns.
No pagination/result limiting documented. list_pipelines and list_tables could return unbounded results. No limit or offset parameters visible, risking context window exhaustion.
Tool descriptions do not explain dependencies or prerequisites. execute_sql_query mentions 'Use list_tables() and get_table_schema()' but this is buried in the description text, not formalized as a dependency hint.
get_load_table description is vague ('Retrieve metadata about data loaded with dlt'). What metadata? Rows processed? Errors? Load timestamp? Missing specificity.