AI Agents - calling custom code and databricks jobs

[GiriSreerangam]

Hi Everyone

I am building AI Agents; my requirement is to call custom tool logic (which was not possible using a unity catalog function) and databricks jobs. I could not find much of documentation on these scenarios. If someone could share any references to documentation would be of great help.

Thank you

#AI Agents #Custom Tools

[SP_6721]

Hi @GiriSreerangam ,

You can try using agent code tools to define Python functions directly within your agent for any custom logic, including Databricks SDK or REST API calls.
https://docs.databricks.com/aws/en/generative-ai/agent-framework/agent-tool

Also, consider exploring MCP (Model Context Protocol) tools for better discoverability and enterprise integration. They allow you to wrap external or custom logic as API endpoints that your agent can easily invoke.
https://docs.databricks.com/aws/en/generative-ai/mcp

[GiriSreerangam]

thank you, @SP_6721. I have gone through this documentation, it was not detailed enough. I was trying with different approaches, could not find a solution yet. 

[jamesl]

Hi @GiriSreerangam , can you share what you are trying to do with the custom tool?

You might be able to implement with a custom MCP server. Here are some other resources that may help: 

• https://github.com/JustTryAI/databricks-mcp-server
• https://github.com/databrickslabs/mcp/tree/master/examples/custom-server
• https://apps-cookbook.dev/docs/dash/workflows/workflows_run

Let us know if you have more details, but if this and @SP_6721 's response helped, then let us know by clicking the "Accept Solution" button.

-James

[GiriSreerangam]

thank you, James. I haven't come across these github links. Let me see if I can find them useful. I am on vacation for the next 3 days. So, I will try exploring them on my return next week. Will confirm here once I take a look. 

Coming to the requirement, on a high level, I have a requirement of reading structured and unstructured data, from documents from an oracle agile server, load them into databricks, enable NLP queries on the data which is loaded. 

Regards

Giri