MCP Services / MCP Functions

Added in version 7.5.0

MCP Functions is a DataFlux Func feature that supports AI Agents calling functions in Func.

Users can expose some functions in DataFlux Func to AI Agents through MCP Services.

1. Integration Steps

Follow the steps below to connect an MCP client / AI Agent to DataFlux Func's MCP Services.

1.1 Create an Access Token

To access DataFlux Func's MCP Services, you need to create an Access Token and enable the MCP Functions feature.

You can also specify which functions to expose in the rules (e.g., by the value of the category field).

1.2 Permission Rules

Permission rules can be edited manually. They are matched from top to bottom, and the first matching rule determines the permission.

Examples:

Rule (Exact ID Match)	Allowed Calls
category:math	Functions whose category is math
scriptSet:mytest	Functions belonging to Script Set mytest
script:mytest__demo	Functions belonging to Script mytest__demo
func:mytest__demo.func	Function whose ID is mytest__demo.func

Rule (Wildcard ID)	Allowed Calls
scriptSet:myte*	Functions matching Script Set ID wildcard myte*
script:mytest__de*	Functions matching Script ID wildcard mytest__de*
func:mytest__demo.fu*	Functions matching Function ID wildcard mytest__demo.fu*

Rule (Deny)	Denied Calls
!category:math	Functions whose category is math

1.3 Configure the MCP Client

DataFlux Func's MCP Services follow the MCP specification

Before actual integration, it is recommended to use MCP Inspector for debugging.

Add the following configuration in the MCP client / AI Agent:

Configuration Item	Value
URL Address	{DataFlux Func access address}/mcp or /mcp2, /mcp3 For specific differences, see MCP Service Mode below.
Request Header	X-Dff-Access-Token: atk-xxxxx:xxxxx

OpenCodeCodexClaude CodeGeneric JSON

{
  "mcp": {
    "dataflux-func-mcp": {
      "enabled": true,
      "type"   : "remote",
      "url"    : "{DataFlux Func access address}/mcp",
      "headers": {
        "X-Dff-Access-Token": "atk-xxxxx:xxxxx"
      }
    }
  }
}

[mcp_servers.dataflux-func-mcp]
enabled      = true
url          = "{DataFlux Func access address}/mcp"
http_headers = { "X-Dff-Access-Token" = "atk-xxxxx:xxxxx" }

{
  "mcpServers": {
    "dataflux-func-mcp": {
      "type": "http",
      "url": "{DataFlux Func access address}/mcp",
      "headers": {
        "X-Dff-Access-Token": "atk-xxxxx:xxxxx"
      }
    }
  }
}

{
  "mcpServers": {
    "dataflux-func-mcp": {
      "url": "{DataFlux Func access address}/mcp",
      "headers": {
        "X-Dff-Access-Token": "atk-xxxxx:xxxxx"
      }
    }
  }
}

Restart the AI Agent after modifying the configuration

Many AI Agents only load MCP Services configuration at startup. After modifying the MCP Services configuration, if the new configuration does not take effect, restart the AI Agent. For example, when using AI Agent plugins such as Codex in VS Code, you can press Ctrl + P and enter >Reload Window to reload the window.

2. MCP Services Mode

DataFlux Func's MCP Services support 3 modes depending on the access path:

Path	Mode
/mcp	Basic Mode
/mcp2	List-Call Mode
/mcp3	Search-Call Mode (Recommended)

Assume the Access Token configuration is as follows, and the codebase contains the following functions:

!category:test
category:math

2.1 /mcp Basic Mode

Directly lists all available functions as Tools

Note: When the function list changes, the client needs to reconnect to update the list.

2.2 /mcp2 List-Call Mode

Provides only two Tools:

• list-func: List functions
• call-func: Call functions

This enables a dynamic function list, requiring the AI Agent/model to correctly follow the List-Call method for invocation.

2.3 /mcp3 Search-Call Mode

Provides only two Tools:

• search-func: Search functions
• call-func: Call functions

This enables searching for the required functions by keyword, requiring the AI Agent/model to correctly follow the Search-Call method for invocation.

When there are many functions, this can significantly reduce context length.