Skip to main content

Documentation Index

Fetch the complete documentation index at: https://docs.mcpcore.io/llms.txt

Use this file to discover all available pages before exploring further.

​
What are parameters?

Parameters are the inputs your tool receives from the AI. When the AI decides to call your tool, it constructs a JSON payload matching your parameter schema. MCPCore validates this payload before your code runs. Your code accesses parameters through the params object: 
// If you defined a parameter named "userId":
const id = params.userId;

​
Adding parameters

In the tool editor, click Add parameter for each input your code needs.

​
Parameter fields

FieldRequiredDescription
NameYesThe key used to access the value in params.*. Use camelCase.
TypeYesThe expected data type (see types below).
DescriptionYesWhat the parameter represents. The AI uses this to know what value to pass.
Required—If checked, the AI must always provide this parameter.
Default value—Used when the parameter is not required and the AI doesn’t provide it.

​
Supported types

​
string

A plain text value. 
// Parameter: "repo", type: string
const repoName = params.repo;      // "anthropics/anthropic-sdk-python"

​
number

An integer or floating-point number. 
// Parameter: "limit", type: number, default: 10
const limit = params.limit;        // 25

​
boolean

true or false. 
// Parameter: "includeArchived", type: boolean, default: false
const includeArchived = params.includeArchived;   // false

​
array

A JSON array. Use the Items type field to specify the type of each element. 
// Parameter: "userIds", type: array of strings
const ids = params.userIds;    // ["user_1", "user_2", "user_3"]

​
object

A JSON object with sub-properties. Define each sub-property with its own type and description. 
// Parameter: "filter" of type object with sub-fields: status (string), since (string)
const { status, since } = params.filter;

​
Best practices for parameter descriptions

The AI reads the parameter description to understand what value to pass. Good descriptions reduce tool call errors.
ParameterWeak descriptionStrong description
repo”The repo""GitHub repository in owner/repo format, e.g. torvalds/linux”
since”Start date""ISO 8601 date string. Filters results to records created after this date, e.g. 2024-01-01”
limit”How many""Maximum number of results to return. Min 1, max 100. Defaults to 20.”

​
Parameter validation

MCPCore validates every tool call against your parameter schema before running the code. If validation fails:
  • Missing required parameters → 400 error returned to the AI
  • Wrong type → 400 error with a clear message
  • Your code is never executed
This means you do not need to validate required parameters in your code — they are guaranteed to be present and of the correct type when your code runs.

​
Optional parameters

For optional parameters, always provide a sensible default value or handle the undefined case in code: 
// Parameter "status" is optional, no default defined
const status = params.status ?? "active";   // default to "active" in code

const rows = await db.select("orders", {
  user_id: params.userId,
  status,
});

​
Working with objects

Object parameters are ideal for grouping related inputs: 
// Parameter "pagination" of type object:
//   - page (number, required)
//   - pageSize (number, default: 20)
const { page, pageSize = 20 } = params.pagination;

const offset = (page - 1) * pageSize;
const rows = await db.query(
  "SELECT * FROM products ORDER BY created_at DESC LIMIT $1 OFFSET $2",
  [pageSize, offset]
);

return { page, pageSize, total: rows.rowCount, results: rows.rows };

​
Working with arrays

// Parameter "tags" of type array of strings
const tags = params.tags;   // ["javascript", "react", "nextjs"]

const results = await db.query(
  "SELECT * FROM posts WHERE tags && $1",
  [tags]
);

return { posts: results.rows };