Create custom Cortex MCP server tools

Create custom OpenAPI and Python tools for Cortex MCP server workflows in Cortex AgentiX.

You can build your own tools using OpenAPI or Python to manage cases, handle issues, and conduct investigations. More detailed information can be found in the README file located in the src/usecase directory. Tools are based on Cortex API endpoints.

Any new or updated components provided by Cortex are automatically downloaded into the remote_components folder. During each update, the folder is fully replaced and all existing contents are recreated. Do not add custom tools to this directory, as it is managed entirely by Cortex and is overwritten at every update.

Create custom Cortex MCP server tools with OpenAPI

You can create an OpenAPI specification for a specific API endpoint.

1. Create a YAML file in the /custom_components/openapi directory with the name of the MCP component. For example: custom_cortex_component.yaml.
2. Base your custom OpenAPI component on the Cortex API documentation structure for a specific endpoint. We recommend viewing the built-in tools, located at /builtin_components/openapi, as a reference.
3. After you define the OpenAPI specification, the Cortex MCP server collects it automatically and it is ready for use.
4. Test your new MCP component by running the Cortex MCP server and writing a prompt that uses your new component.

Create custom Cortex MCP server tools with Python

We recommend using Python for more complex MCP components that require custom logic. MCP components in Python are defined in a module.

1. Create a new Python file in the /custom_components directory.
2. Define a class that inherits from the BaseModule class with the required methods. We recommend viewing the built-in modules, located at /builtin_components, as a reference.
3. After you define a class, the Cortex MCP server collects it automatically and it is ready for use.
4. Test your new MCP component by adding an end-to-end test in the tests/e2e directory or run the MCP server and write a prompt that uses your new component.