Bring the ability to deliver "custom" tools directly in database and let the MCP server dynamically load it (no fiddling with config files and server restart/testing needed, complete transparency and control on the logic implemented from the database).
This addresses one of our key requirement which is "how do I restrict the data scope and possible queries to provide tools that are relevant and accurate in a specific business application?"
Today we allow this via the "Custom Tools" concept, which is server-side configuration and as such limits the developer experience: creating a new tool is a server admin duty when it should be a data engineer's.
Making tools "yet another database object" also simplify governance and lineage...
Continue the work on the https://github.com/Teradata/teradata-mcp-server/tree/db-tool-registry branch:
[ ] Find better strategy for "initial load" of tools (load with the first database connection rather than at the server startup - blocking)
[ ] Add support for "Table" objects (tables/views) with "cube-like" features (ie. current implementation. Ie. filter and group operations - try to simplify from current align to typical table operations/pivot).
[ ] Add macros to do tools registrations and checking.
[ ] Add a STO or ChatCompletion example
Bring the ability to deliver "custom" tools directly in database and let the MCP server dynamically load it (no fiddling with config files and server restart/testing needed, complete transparency and control on the logic implemented from the database).
This addresses one of our key requirement which is "how do I restrict the data scope and possible queries to provide tools that are relevant and accurate in a specific business application?"
Today we allow this via the "Custom Tools" concept, which is server-side configuration and as such limits the developer experience: creating a new tool is a server admin duty when it should be a data engineer's.
Making tools "yet another database object" also simplify governance and lineage...
Continue the work on the https://github.com/Teradata/teradata-mcp-server/tree/db-tool-registry branch:
[ ] Find better strategy for "initial load" of tools (load with the first database connection rather than at the server startup - blocking)
[ ] Add support for "Table" objects (tables/views) with "cube-like" features (ie. current implementation. Ie. filter and group operations - try to simplify from current align to typical table operations/pivot).
[ ] Add macros to do tools registrations and checking.
[ ] Add a STO or ChatCompletion example