AI GitHub
Semantic Kernel for integrating models into applications
Expose application functions to models through plugins and provider connectors.
Key Takeaways
- Semantic Kernel organizes callable functions and model connections inside an application framework.
- Use a supported SDK and map one existing operation into a plugin.
- Wrap a read-only business function and test its inputs with both a direct call and a model-mediated call.
- A plugin description is part of the interface the model sees.
How it works
Semantic Kernel organizes callable functions and model connections inside an application framework. A plugin can make existing domain operations available with a model-facing description.
What you need
Use a supported SDK and map one existing operation into a plugin. Define what arguments the underlying service accepts and which caller identity owns the operation.
What to try first
Wrap a read-only business function and test its inputs with both a direct call and a model-mediated call. Inspect what the model receives when the function returns an error.
Test an unfamiliar argument passed to the plugin. Well-defined errors make it easier for both application code and a model to recover.
Where to look for failures
A plugin description is part of the interface the model sees. Keep model-facing operations narrow, verify SDK support for your language and put permissions in the underlying application.
Related Franklin pages
Build the same two-step workflow with one tool and one failure case. Inspect how state, errors and tool permissions cross the boundary between the steps.
- Create agents and workflows with tools, session state and development utilities. Google ADK for Python agent applications
- Represent an agent workflow as nodes, transitions and explicit state. LangGraph for stateful agent workflows
Browse Franklin for concrete examples of tools and services exposed through MCP. Franklin’s MCP directory
Project documentation
Try it for your workflow
Explore the project’s documentation for setup and usage, or save this page to a collection for later.
Read the project documentation ↗