More screenshots

    Works with the AI Agents you already use

    Claude CodeClaude CodeCursorCursorCodex CLICodex CLIMuseMuseDotsDots+21 more

    mcp gateway

    1

    Orchestrate multiple MCP servers and tools through a single gateway to bypass function limits and context bloat.

    $9.99

    Secure checkout via Stripe

    Money-back guarantee

    Converts to your local currency at checkout

    Security scanned
    mcp gateway

    mcp gateway

    Example session with this skill installed

    I need to add a new GitHub MCP server using npx and my GITHUB_TOKEN variable. Can you set that up in the gateway for me?

    • Read your context and instructions
    • Compiled the mcp gateway

    Managing multiple Model Context Protocol (MCP) servers often hits a wall when tool counts exceed the context window or LLM function declaration limits. This skill teaches AI coding agents how to install, configure, and orchestrate the open-source

    MCP Gateway runtime (https://github.com/fathasss/mcp-gateway, MIT License) as a single entry point for your MCP tools.

    What it does

    Dynamic Routing: Keeps only 5 core orchestrator tools loaded and routes tool discovery (search_tools) and execution (call_tool) on demand.

    Consent-First Client Migration: Previews existing configurations from Claude Code, Cursor, Claude Desktop, or Gemini CLI (dry-run + timestamped backup) and migrates them into the gateway after explicit user approval.

    Server Orchestration: Adds, updates, enables, disables, or removes downstream stdio, http, streamableHttp, and sse MCP servers via manage_mcp_server without manual JSON editing.

    On-Demand Discovery: Searches a local indexed tool cache (tools-cache.json) with English/Turkish synonym expansion when needed.

    Secret Extraction & Redaction: Extracts sensitive keys into a local .env file with ${VAR_NAME} placeholders in config.json, and redacts Gateway config outputs and Gateway error logs.

    How it works

    Install & Connect (with approval): Checks prerequisites (Node.js >= 18, Git), clones/installs mcp-gateway, and safely migrates or registers your client config after explicit confirmation.
    2.

    Discover: Queries search_tools to retrieve only the relevant tool names and inputSchema definitions for the current task.
    3. Execute: Invokes downstream tools through call_tool using the target server ID, tool name, and arguments.
    4.

    Manage: Uses manage_mcp_server and list_mcp_servers to hot-plug or toggle servers without restarting your IDE.

    Frameworks & tools

    Built for the Model Context Protocol (MCP) ecosystem. Requires Node.js >= 18.0.0, Git, and the open-source mcp-gateway server; integrates with Claude Code, Cursor, Claude Desktop, and Gemini CLI.

    Why this beats prompting it yourself

    Connecting multiple MCP servers directly injects dozens or hundreds of tool schemas into every prompt turn, quickly hitting 64 or 128 function declaration caps. Instead of loading all downstream schemas upfront, this workflow exposes 5 core Gateway tools and fetches only the schemas needed for the active sub-task.

    Use cases

    • Consolidate 5–15+ MCP servers behind one gateway entry to stay well below 64/128 function declaration caps.
    • Safely migrate an existing Claude Code, Cursor, Claude Desktop, or Gemini CLI mcpServers config into mcp-gateway with automatic backups and secret extraction to .env.
    • Hot-plug or toggle local stdio (npx, uvx, Docker) or StreamableHTTP/SSE MCP servers without restarting your editor.
    • Discover a needed capability across many servers (e.g., "create pull request") without memorizing exact server/tool names.

    Known limitations

    Does not handle interactive browser OAuth PKCE flows headlessly without a valid token in .env. Claude Desktop lacks native directory-based SKILL.md auto-discovery. Automated --client migration presets cover only Claude Code, Cursor, Claude Desktop, and Gemini CLI. Credentials in .env are stored as local plaintext, and third-party call_tool response payloads are passed through without Gateway redaction.

    Connects securely to your tools. The creator never sees your data.

    About this skill

    Managing multiple Model Context Protocol (MCP) servers often hits a wall when tool counts exceed the context window or function calling limits of coding agents. This skill implements a dynamic gateway that acts as a single point of entry for all your MCP tools.

    What it does

    • Dynamic Routing routes tool discovery and execution through a two-phase process to keep the agent context window clean.
    • Client Migration handles the move of existing configurations from Claude Code, Cursor, Claude Desktop, or Gemini CLI into the gateway.
    • Server Orchestration allows for adding, removing, enabling, or disabling MCP servers via stdio or http/sse transports without manual file editing.
    • On-demand Discovery searches a local index of tools to find specific capabilities only when they are needed for a task.
    • Secret Hygiene manages sensitive environment variables via a local .env file and placeholders to prevent token leakage in chat logs.

    How it works

    1. Connect the gateway to your preferred coding agent by running the migration script to move existing MCP configs.
    2. Discover available tools using search_tools with a natural language query to identify the correct server and function name.
    3. Execute downstream tasks by passing the server ID and arguments to the call_tool interface.
    4. Manage the lifecycle of connected servers using manage_mcp_server to hot-plug new capabilities or disable unused ones.

    Frameworks & tools

    Built for the Model Context Protocol (MCP) ecosystem. Supports Node.js environments and integrates with Claude Code, Cursor, Claude Desktop, and Gemini CLI.

    Why this beats prompting it yourself

    Standard agents attempt to load every tool from every server into every turn, quickly hitting the 64 or 128 function limit. This skill abstracts that complexity, allowing you to connect dozens of servers while only exposing the relevant tools for the current sub-task.

    Use cases

    • Consolidate five different MCP servers into a single agent interface without hitting function caps.
    • Migrate a complex Claude Desktop config to Claude Code while preserving all environment variables and paths.
    • Add or test new MCP servers via HTTP or stdio without restarting your IDE or editing JSON files.
    • Search for a specific capability (e.g., "create jira issue") across multiple disparate toolsets.

    Known limitations

    Does not support headless OAuth PKCE flows for remote servers. Claude Desktop lacks native auto-discovery for SKILL.md directories. Automatic migration is restricted to the four supported clients.

    Requires Node.js 18+. Auto-migration supports Claude Code, Cursor, Claude Desktop, and Gemini CLI only. Headless OAuth PKCE flows are not supported.

    How to use

    1. 1

      Connect your agent to the Agensi MCP

      Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup

    2. 2

      Sign in to Agensi

      Your agent shows a sign-in link. Sign in and click Allow.

    3. 3

      Ask for what you need

      Your agent loads this skill and uses it on your task. It always asks before any payment.

    Reviews

    No reviews yet

    Be one of the first to try it. Every listed skill goes through the trust checks below.

    Security scanned

    Checked by our 8-point scan before listing

    Fresh listing

    Recently published to Agensi

    Money-back guarantee

    Doesn't do what its page promises? You get your money back.

    Trust & safety

    Security scan

    Security scan passed 4 days ago

    Requires

    GitHub
    Gateway MCP
    Node.js
    Git

    Permissions

    Terminal / Shell
    Write Files
    Read Files
    Network Access
    Environment Variables

    Allowed Hosts

    api.github.com
    Listed4 days ago

    What's inside

    Frequently Asked Questions

    The gateway solves the "context bloat" problem where loading too many MCP servers simultaneously exceeds an agent's function calling limits. It acts as a single interface, dynamically indexing your tools so only the relevant ones are injected into the active chat session.

    The skill includes a migration script that automatically imports existing configurations from Claude Code, Cursor, Claude Desktop, and Gemini CLI. Once migrated, you can manage all servers via the gateway instead of manually editing individual JSON config files.

    The gateway supports both `stdio` and `http/sse` transports, making it compatible with the vast majority of local and remote MCP servers. It is specifically optimized for Node.js environments and leading coding agents like Claude Code and Cursor.

    No, the gateway features "hot-plug" server management. You can add, remove, enable, or disable servers using the `manage_mcp_server` tool through the agent interface without needing to restart your IDE or manually kill processes.

    The skill uses a dedicated local `.env` file and placeholders to store sensitive credentials. This ensures your API keys and tokens remain in your local environment and are not leaked into the LLM's chat logs or prompt history.

    You receive the full MCP server package, the migration utility script, and documentation for integrating the gateway with supported IDEs and CLI agents. This includes the logic for the two-phase tool discovery and execution process.

    New in Agents & Orchestration