Dynamic Inputs

[agardnerIT]

Pre-submission Checklist

• I have checked that this question would not be more appropriate as an issue in a specific repository
• I have searched existing discussions and documentation for answers

Question Category

• Protocol Specification
• SDK Usage
• Server Implementation
• General Implementation
• Documentation
• Other

Your Question

I am new to MCP so feel free to challenge / correct anything here!

I have a third party API that accepts POST body like:

get logs
| filter contains(body, "someText")
| filter contains(body, "someOtherText")

So I write a tool called get_logs

This would retrieve only logs where the body contains someText and someOtherText.

The model is smart enough to figure it needs to run the get_logs tool but how do I dynamically parameterise the filters?
Note that | filter contains(...) is just one option. It could be | filter matchesPhrase(attributes["foo"], "bar") to return structured logs where the foo attribute == bar

I can't figure out how to do that dynamic parameterisation (Python in particular). Maybe there's a different way?

[andrewlayer]

I am not sure I completely understand your question, but from what I can tell it looks like you are looking for Enums. You want the LLM to give specific structures of arguments. This might be challenging because you need to outline each and every way you want to allow the LLM to call a function. Either way, the solution to your problem lies in figuring out what you want in the JSON schema for your tool body. Here you can see the JSON schema spec. Also I provided an example below using enums in json schema.

{
  "properties": {
    "filters": {
      "type": "array",
      "items": {
        "type": "object",
        "properties": {
          "type": {
            "type": "string",
            "enum": ["contains", "matchesPhrase", "equals", "startsWith", "endsWith", "regex"],
            "description": "The type of filter operation to perform"
          },
          "field": {
            "type": "string",
            "description": "The field to filter on (e.g., 'body' or 'attributes.foo')"
          },
          "value": {
            "type": ["string", "number", "boolean"],
            "description": "The value to filter against"
          }
        },
        "required": ["type", "field", "value"],
        "additionalProperties": false
      },
      "description": "Array of filter operations to apply to the logs"
    }
  },
  "type": "object",
  "required": ["filters"],
  "examples": [
    {
      "filters": [
        {
          "type": "contains",
          "field": "body",
          "value": "someText"
        },
        {
          "type": "matchesPhrase",
          "field": "attributes.foo",
          "value": "bar"
        }
      ]
    }
  ]
}

[honghuizheng]

I have the same issue。i split the filters by regex in my tools