Unified Gateway

Open-source toolkit

A single point of connection for all your tools. Reduce complexity and improve performance by loading tool contexts only when they are needed.

01 — Overview

Most setups load every tool definition at once, consuming thousands of tokens before any work begins. The Unified Gateway solves this by exposing only core capabilities initially and loading detailed context on-demand — reducing initial overhead by more than 90%.

💡 Toolkit boundary:

This page describes the open-source aicode-toolkit gateway. Its meta-tools are describe_tools and use_tool. Agiflow platform assistant connections use describe_capabilities, use_capability, and get_current_scope.

02 — Core Benefits

Token Efficiency

Instead of overwhelming your assistants with thousands of tool descriptions, the gateway provides a clean initial state. Resources are requested progressively as the task requires.

Centralized Control

Connect multiple sources of knowledge and automation through a single endpoint. Replace fragmented configurations with one unified board.

Controlled Access

Define which tools are available for different tasks or team roles. Maintain strict control over what your AI assistants can access and execute.

Flexible Setup

Support for both local and remote connection types. Mix and match tool sources while maintaining a consistent interface for your AI team.

03 — Core Capabilities

describe_tools

Retrieve information about available capabilities from all connected sources. Supports focused filtering by source name or task type.

List all sources
Filter by capability
Get detailed schemas

use_tool

Execute any discovered capability through the gateway. Includes automatic routing to the correct source and validation of arguments.

Standardized execution
Automatic routing
Structured results

04 — Setup

Quick start

Initialize the gateway configuration in your local project:

npx @agiflowai/one-mcp init

Environment configuration

Add this to your AI assistant's connection settings:

{
  "mcpServers": {
    "gateway": {
      "command": "npx",
      "args": [
        "-y",
        "@agiflowai/one-mcp",
        "mcp-serve",
        "--config",
        "./mcp-config.yaml"
      ]
    }
  }
}

Source definitions

Define your knowledge sources in a central configuration file:

mcpServers:
  # Local file system
  filesystem:
    command: npx
    args: ["-y", "@modelcontextprotocol/server-filesystem", "${HOME}/Documents"]
    config:
      instruction: "Secure file access"
      toolBlacklist:
        - write_file
        - delete_file

  # Remote company tools
  company-tools:
    url: "https://mcp.company.com/sse"
    transport: sse
    config:
      omitToolDescription: true

  # Experimental features (disabled)
  experimental:
    command: npx
    args: ["-y", "experimental-tools"]
    disabled: true

💡 What happens if a tool source is unavailable?:

The gateway skips any source it cannot reach and continues serving tools from the remaining sources. Your assistant still works — it just won’t see the tools from the unavailable source until it comes back online.

05 — Terminal Commands

mcp-serve

Start the gateway server for local or remote connections.

init

Initialize a new configuration file in your directory.

list-tools

Display all capabilities currently shared through the gateway.

describe-tools

Get detailed information about specific capabilities.

use-tool

Execute a capability directly from your terminal.

06 — Related

Standardization

Generate consistent results using shared templates for your projects.

View Source

Explore the open-source toolkit, contribute, or report issues on GitHub.