Best Practices for MCP Server Development (Preview)
This topic provides information and guidelines to help you with your Oracle Analytics Cloud MCP Server development tasks.
Recommended Workflow
| Step | Action | Instructions |
|---|---|---|
| 1 | Discover | Call discover_data to list
available data sources.
|
| 2 | Describe | Call describe_data with
tablesOnly: true to get table list.
|
| 3 | Inspect | Call describe_data for specific
tables to get columns.
|
| 4 | Query | Build and execute Logical SQL query. |
Performance Guidelines
| Guideline | Best Practice |
|---|---|
| Fetch Size | Start with 1000 and increase as needed. |
| Time Filters | Always filter by time dimension. |
| Column Selection | Request only needed columns. |
| TOPN versus ORDER BY | Use TOPN for filtering. Use ORDER BY for display. |
Development Checklist
Use this checklist to create valid queries and prevent build errors and issues:
Make sure that:
- Columns use three-part fully qualified names.
- All columns have
llm_prefixed aliases. - WHERE clause includes a time dimension filter.
- ORDER BY present when using FETCH.
- FETCH FIRST n ROWS ONLY included.
- No manual JOIN syntax.
- OVERRIDEAGGR is used when overriding default aggregation.