Skip to main content
Glama
gendosu

Notion MCP Server

by gendosu

API-get-block-children

Retrieve child blocks from a specific block in Notion using the block ID. Supports pagination with start cursor and page size for efficient data access.

Instructions

Notion | Retrieve block children

Input Schema

TableJSON Schema
NameRequiredDescriptionDefault
block_idYesIdentifier for a [block](ref:block)
page_sizeNoThe number of items from the full list desired in the response. Maximum: 100
start_cursorNoIf supplied, this endpoint will return a page of results starting after the cursor provided. If not supplied, this endpoint will return the first page of results.

Implementation Reference

  • MCP callTool request handler: locates the OpenAPI operation corresponding to the tool name 'API-get-block-children' (maps to getBlockChildren operationId) and executes it via HTTP client, returning the JSON response or structured error.
    this.server.setRequestHandler(CallToolRequestSchema, async (request) => {
      const { name, arguments: params } = request.params
    
      // Find the operation in OpenAPI spec
      const operation = this.findOperation(name)
      if (!operation) {
        throw new Error(`Method ${name} not found`)
      }
    
      try {
        // Execute the operation
        const response = await this.httpClient.executeOperation(operation, params)
    
        // Convert response to MCP format
        return {
          content: [
            {
              type: 'text', // currently this is the only type that seems to be used by mcp server
              text: JSON.stringify(response.data), // TODO: pass through the http status code text?
            },
          ],
        }
      } catch (error) {
        console.error('Error in tool call', error)
        if (error instanceof HttpClientError) {
          console.error('HttpClientError encountered, returning structured error', error)
          const data = error.data?.response?.data ?? error.data ?? {}
          return {
            content: [
              {
                type: 'text',
                text: JSON.stringify({
                  status: 'error', // TODO: get this from http status code?
                  ...(typeof data === 'object' ? data : { data: data }),
                }),
              },
            ],
          }
        }
        throw error
      }
    })
  • Executes the HTTP request for the OpenAPI operation corresponding to the tool, handling parameters, path/query separation, file uploads, and error handling.
    async executeOperation<T = any>(
      operation: OpenAPIV3.OperationObject & { method: string; path: string },
      params: Record<string, any> = {},
    ): Promise<HttpClientResponse<T>> {
      const api = await this.api
      const operationId = operation.operationId
      if (!operationId) {
        throw new Error('Operation ID is required')
      }
    
      // Handle file uploads if present
      const formData = await this.prepareFileUpload(operation, params)
    
      // Separate parameters based on their location
      const urlParameters: Record<string, any> = {}
      const bodyParams: Record<string, any> = formData || { ...params }
    
      // Extract path and query parameters based on operation definition
      if (operation.parameters) {
        for (const param of operation.parameters) {
          if ('name' in param && param.name && param.in) {
            if (param.in === 'path' || param.in === 'query') {
              if (params[param.name] !== undefined) {
                urlParameters[param.name] = params[param.name]
                if (!formData) {
                  delete bodyParams[param.name]
                }
              }
            }
          }
        }
      }
    
      // Add all parameters as url parameters if there is no requestBody defined
      if (!operation.requestBody && !formData) {
        for (const key in bodyParams) {
          if (bodyParams[key] !== undefined) {
            urlParameters[key] = bodyParams[key]
            delete bodyParams[key]
          }
        }
      }
    
      const operationFn = (api as any)[operationId]
      if (!operationFn) {
        throw new Error(`Operation ${operationId} not found`)
      }
    
      try {
        // If we have form data, we need to set the correct headers
        const hasBody = Object.keys(bodyParams).length > 0
        const headers = formData
          ? formData.getHeaders()
          : { ...(hasBody ? { 'Content-Type': 'application/json' } : { 'Content-Type': null }) }
        const requestConfig = {
          headers: {
            ...headers,
          },
        }
    
        // first argument is url parameters, second is body parameters
        const response = await operationFn(urlParameters, hasBody ? bodyParams : undefined, requestConfig)
    
        // Convert axios headers to Headers object
        const responseHeaders = new Headers()
        Object.entries(response.headers).forEach(([key, value]) => {
          if (value) responseHeaders.append(key, value.toString())
        })
    
        return {
          data: response.data,
          status: response.status,
          headers: responseHeaders,
        }
      } catch (error: any) {
        if (error.response) {
          console.error('Error in http client', error)
          const headers = new Headers()
          Object.entries(error.response.headers).forEach(([key, value]) => {
            if (value) headers.append(key, value.toString())
          })
    
          throw new HttpClientError(error.response.statusText || 'Request failed', error.response.status, error.response.data, headers)
        }
        throw error
      }
  • Generates the inputSchema (JSON Schema) for each MCP tool method from the OpenAPI operation's parameters and requestBody, including handling refs, descriptions, required fields, and file uploads.
    private convertOperationToMCPMethod(operation: OpenAPIV3.OperationObject, method: string, path: string): NewToolMethod | null {
      if (!operation.operationId) {
        console.warn(`Operation without operationId at ${method} ${path}`)
        return null
      }
    
      const methodName = operation.operationId
    
      const inputSchema: IJsonSchema & { type: 'object' } = {
        $defs: this.convertComponentsToJsonSchema(),
        type: 'object',
        properties: {},
        required: [],
      }
    
      // Handle parameters (path, query, header, cookie)
      if (operation.parameters) {
        for (const param of operation.parameters) {
          const paramObj = this.resolveParameter(param)
          if (paramObj && paramObj.schema) {
            const schema = this.convertOpenApiSchemaToJsonSchema(paramObj.schema, new Set(), false)
            // Merge parameter-level description if available
            if (paramObj.description) {
              schema.description = paramObj.description
            }
            inputSchema.properties![paramObj.name] = schema
            if (paramObj.required) {
              inputSchema.required!.push(paramObj.name)
            }
          }
        }
      }
    
      // Handle requestBody
      if (operation.requestBody) {
        const bodyObj = this.resolveRequestBody(operation.requestBody)
        if (bodyObj?.content) {
          // Handle multipart/form-data for file uploads
          // We convert the multipart/form-data schema to a JSON schema and we require
          // that the user passes in a string for each file that points to the local file
          if (bodyObj.content['multipart/form-data']?.schema) {
            const formSchema = this.convertOpenApiSchemaToJsonSchema(bodyObj.content['multipart/form-data'].schema, new Set(), false)
            if (formSchema.type === 'object' && formSchema.properties) {
              for (const [name, propSchema] of Object.entries(formSchema.properties)) {
                inputSchema.properties![name] = propSchema
              }
              if (formSchema.required) {
                inputSchema.required!.push(...formSchema.required!)
              }
            }
          }
          // Handle application/json
          else if (bodyObj.content['application/json']?.schema) {
            const bodySchema = this.convertOpenApiSchemaToJsonSchema(bodyObj.content['application/json'].schema, new Set(), false)
            // Merge body schema into the inputSchema's properties
            if (bodySchema.type === 'object' && bodySchema.properties) {
              for (const [name, propSchema] of Object.entries(bodySchema.properties)) {
                inputSchema.properties![name] = propSchema
              }
              if (bodySchema.required) {
                inputSchema.required!.push(...bodySchema.required!)
              }
            } else {
              // If the request body is not an object, just put it under "body"
              inputSchema.properties!['body'] = bodySchema
              inputSchema.required!.push('body')
            }
          }
        }
      }
    
      // Build description including error responses
      let description = operation.summary || operation.description || ''
      if (operation.responses) {
        const errorResponses = Object.entries(operation.responses)
          .filter(([code]) => code.startsWith('4') || code.startsWith('5'))
          .map(([code, response]) => {
            const responseObj = this.resolveResponse(response)
            let errorDesc = responseObj?.description || ''
            return `${code}: ${errorDesc}`
          })
    
        if (errorResponses.length > 0) {
          description += '\nError Responses:\n' + errorResponses.join('\n')
        }
      }
    
      // Extract return type (response schema)
      const returnSchema = this.extractResponseType(operation.responses)
    
      // Generate Zod schema from input schema
      try {
        // const zodSchemaStr = jsonSchemaToZod(inputSchema, { module: "cjs" })
        // console.log(zodSchemaStr)
        // // Execute the function with the zod instance
        // const zodSchema = eval(zodSchemaStr) as z.ZodType
    
        return {
          name: methodName,
          description,
          inputSchema,
          ...(returnSchema ? { returnSchema } : {}),
        }
      } catch (error) {
        console.warn(`Failed to generate Zod schema for ${methodName}:`, error)
        // Fallback to a basic object schema
        return {
          name: methodName,
          description,
          inputSchema,
          ...(returnSchema ? { returnSchema } : {}),
        }
      }
    }
  • Registers the MCP listTools request handler, which dynamically constructs and returns the list of available tools (including 'API-get-block-children') from parsed OpenAPI operations.
    // Handle tool listing
    this.server.setRequestHandler(ListToolsRequestSchema, async () => {
      const tools: Tool[] = []
    
      // Add methods as separate tools to match the MCP format
      Object.entries(this.tools).forEach(([toolName, def]) => {
        def.methods.forEach(method => {
          const toolNameWithMethod = `${toolName}-${method.name}`;
          const truncatedToolName = this.truncateToolName(toolNameWithMethod);
          tools.push({
            name: truncatedToolName,
            description: method.description,
            inputSchema: method.inputSchema as Tool['inputSchema'],
          })
        })
      })
    
      return { tools }
    })
  • Parses OpenAPI spec into MCP tools under 'API' group; creates tool names as 'API-{operationId}' (e.g., 'API-getBlockChildren' or 'API-get-block-children'), inputSchemas, and lookup map for dispatch.
    convertToMCPTools(): {
      tools: Record<string, { methods: NewToolMethod[] }>
      openApiLookup: Record<string, OpenAPIV3.OperationObject & { method: string; path: string }>
      zip: Record<string, { openApi: OpenAPIV3.OperationObject & { method: string; path: string }; mcp: NewToolMethod }>
    } {
      const apiName = 'API'
    
      const openApiLookup: Record<string, OpenAPIV3.OperationObject & { method: string; path: string }> = {}
      const tools: Record<string, { methods: NewToolMethod[] }> = {
        [apiName]: { methods: [] },
      }
      const zip: Record<string, { openApi: OpenAPIV3.OperationObject & { method: string; path: string }; mcp: NewToolMethod }> = {}
      for (const [path, pathItem] of Object.entries(this.openApiSpec.paths || {})) {
        if (!pathItem) continue
    
        for (const [method, operation] of Object.entries(pathItem)) {
          if (!this.isOperation(method, operation)) continue
    
          const mcpMethod = this.convertOperationToMCPMethod(operation, method, path)
          if (mcpMethod) {
            const uniqueName = this.ensureUniqueName(mcpMethod.name)
            mcpMethod.name = uniqueName
            mcpMethod.description = this.getDescription(operation.summary || operation.description || '')
            tools[apiName]!.methods.push(mcpMethod)
            openApiLookup[apiName + '-' + uniqueName] = { ...operation, method, path }
            zip[apiName + '-' + uniqueName] = { openApi: { ...operation, method, path }, mcp: mcpMethod }
          }
        }
      }
    
      return { tools, openApiLookup, zip }
    }

Schema Changelog

Changes observed during successful MCP inspections.

  1. First observedv1.0.0

TDQS

C2.9/5.0
Behavior2/5

Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?

No annotations are provided, so the description carries the full burden. It states 'Retrieve' implying a read-only operation, but lacks details on permissions, rate limits, error handling, or what 'block children' entails (e.g., nested content). The description is minimal and doesn't disclose behavioral traits beyond the basic action.

Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.

Conciseness5/5

Is the description appropriately sized, front-loaded, and free of redundancy?

The description is extremely concise ('Notion | Retrieve block children') with no wasted words. It front-loads the key information (context and action) efficiently, though it may be overly brief for a tool with three parameters and no annotations.

Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.

Completeness2/5

Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?

Given the tool has 3 parameters, no annotations, and no output schema, the description is incomplete. It doesn't explain the return values (e.g., list of child blocks), pagination behavior, or error cases. For a retrieval tool with moderate complexity, more context is needed to guide effective use.

Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.

Parameters3/5

Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?

Schema description coverage is 100%, with clear documentation for block_id, page_size, and start_cursor. The description adds no additional parameter semantics beyond what the schema provides, such as examples or context for 'block children'. This meets the baseline for high schema coverage.

Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.

Purpose4/5

Does the description clearly state what the tool does and how it differs from similar tools?

The description clearly states the action ('Retrieve') and resource ('block children') with the Notion context. It distinguishes this as a read operation (vs. create/update/delete siblings like API-create-a-database or API-delete-a-block), though it doesn't explicitly differentiate from similar retrieval tools like API-retrieve-a-block or API-retrieve-a-page.

Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.

Usage Guidelines2/5

Does the description explain when to use this tool, when not to, or what alternatives exist?

No guidance is provided on when to use this tool versus alternatives. It doesn't mention when to retrieve block children versus retrieving a block directly (API-retrieve-a-block) or searching (API-post-search), nor does it specify prerequisites like needing a valid block_id or handling pagination.

Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.