MCP Access Point Details

MCP Access Point is a lightweight gateway that turns existing HTTP services into MCP (Model Context Protocol) endpoints with zero code changes. Built on high-performance Pingora proxy, it enables seamless protocol conversion between HTTP and MCP, supporting both SSE and Streamable HTTP. Designed for multi-tenant deployments, it offers a RESTful Admin API for real-time configuration management, dynamic updates, and resource administration without restarting the service. This repository provides a clear Quick Start, multi-tenancy guidance, and admin operations to manage upstreams, services, routes, and more, making it easy to expose legacy HTTP APIs to MCP clients like Cursor Desktop and MCP Inspectors.

Use Case

Use MCP Access Point to bridge existing HTTP services into the MCP ecosystem. It supports multi-tenancy, dynamic runtime configuration, and admin-controlled resource management. Example use case: you have several internal HTTP microservices and want to expose them to MCP clients without touching code. Define mcps and upstreams in config.yaml, start the gateway, and optionally manage configurations via the Admin API. Key workflow includes configuring routes with operation_id, such as get_weather, to map MCP operations to backend HTTP endpoints. Example from the docs: - operation_id: get_weather - uri: /points/{latitude},{longitude} - method: GET - meta includes name: Get Weather and description with inputSchema enforcing latitude and longitude ranges.

Available Tools (1)

Examples & Tutorials

From the documentation:

Config snippet showing a route using operation_id get_weather:

<h1 class="text-2xl font-semibold mt-5 mb-3">config.yaml example (supports multiple services)</h1>
mcps:
  • id: service-1 # Unique identifier, accessible via /api/service-1/sse or /api/service-1/mcp

  • upstream_id: 1
    path: config/openapi_for_demo_patch1.json # Local OpenAPI spec path
  • id: service-2 # Unique identifier

  • upstream_id: 2
    path: https://petstore.swagger.io/v2/swagger.json # Remote OpenAPI spec
  • id: service-3

  • upstream_id: 3
    routes: # Custom routing
  • id: 1

  • operation_id: get_weather
    uri: /points/{latitude},{longitude}
    method: GET
    meta:
    name: Get Weather
    description: Retrieve weather information by coordinates
    inputSchema: # Optional input validation
    type: object
    required:
  • latitude

  • longitude

  • properties:
    latitude:
    type: number
    minimum: -90
    maximum: 90
    longitude:
    type: number
    minimum: -180
    maximum: 180

    upstreams: # Required upstream configuration

  • id: 1

  • headers: # Headers to send to upstream service
    X-API-Key: "12345-abcdef" # API key
    Authorization: "Bearer token123" # Bearer token
    User-Agent: "MyApp/1.0" # User agent
    Accept: "application/json" # Accept header
    nodes: # Backend nodes (IP or domain)
    "127.0.0.1:8090": 1 # Format: address:weight

    To run:

    cargo run -- -c config.yaml

    Quick Start (from docs):

    <h1 class="text-2xl font-semibold mt-5 mb-3">Install from source</h1>
    git clone https://github.com/sxhxliang/mcp-access-point.git
    cd mcp-access-point
    cargo run -- -c config.yaml

    <h1 class="text-2xl font-semibold mt-5 mb-3">Use inspector for debugging (start service first)</h1>
    npx @modelcontextprotocol/inspector node build/index.js
    <h1 class="text-2xl font-semibold mt-5 mb-3">Access http://127.0.0.1:6274/</h1>
    <h1 class="text-2xl font-semibold mt-5 mb-3">Select "SSE" and enter 0.0.0.0:8080/sse, then click connect</h1>
    <h1 class="text-2xl font-semibold mt-5 mb-3">or select "Streamable HTTP" and enter 0.0.0.0:8080/mcp</h1>

    Running via Docker (examples):

    docker run -d --name mcp-access-point --rm \
    -p 8080:8080 \
    -e port=8080 \
    -v /path/to/your/config.yaml:/app/config/config.yaml \
    ghcr.io/sxhxliang/mcp-access-point:main

    Environment variable reference:

    - port: MCP Access Point listening port (default: 8080)

    Installation Guide

    Step-by-step (from the documentation):

    1) Install from source

    <h1 class="text-2xl font-semibold mt-5 mb-3">Install from source</h1>
    git clone https://github.com/sxhxliang/mcp-access-point.git
    cd mcp-access-point
    cargo run -- -c config.yaml

    2) Quick Start commands for debugging with Inspector

    <h1 class="text-2xl font-semibold mt-5 mb-3">Use inspector for debugging (start service first)</h1>
    npx @modelcontextprotocol/inspector node build/index.js
    <h1 class="text-2xl font-semibold mt-5 mb-3">Access http://127.0.0.1:6274/</h1>
    <h1 class="text-2xl font-semibold mt-5 mb-3">Select "SSE" and enter 0.0.0.0:8080/sse, then click connect</h1>
    <h1 class="text-2xl font-semibold mt-5 mb-3">or select "Streamable HTTP" and enter 0.0.0.0:8080/mcp</h1>

    3) Docker run example

    docker run -d --name mcp-access-point --rm \
    -p 8080:8080 \
    -e port=8080 \
    -v /path/to/your/config.yaml:/app/config/config.yaml \
    ghcr.io/sxhxliang/mcp-access-point:main

    4) Docker image build (optional)

    <h1 class="text-2xl font-semibold mt-5 mb-3">Clone repository</h1>
    git clone https://github.com/sxhxliang/mcp-access-point.git
    cd mcp-access-point

    <h1 class="text-2xl font-semibold mt-5 mb-3">Build image</h1>
    docker build -t liangshihua/mcp-access-point:latest .

    5) Environment variable reference

    - port: MCP Access Point listening port (default: 8080)

    Frequently Asked Questions

    Is this your MCP?

    Claim ownership and get verified badge

    Repository Stats

    Sponsored

    Ad Space Available
    Important Notes

    Notes and important points from the documentation:

  • MCP Access Point supports multi-tenancy and dynamic runtime configuration via the Admin API.

  • The Admin API runs on port 8081 by default (see example in Admin API configuration).

  • Admin API endpoints allow resource management, batch operations, and reloading configurations.

  • The default public port for the MCP Access Point is 8080 unless overridden via -e port or in config.yaml.

  • The system supports both SSE and Streamable HTTP modes for MCP clients.

  • Configuration is provided via a config.yaml file (mcps and upstreams sections).

  • You can reload the full configuration from file using the Admin API (POST /admin/reload/config) with optional config_path in the body.

  • The README provides concrete examples for creating resources, batch operations, and getting resource statistics via the Admin API.

  • Prerequisites

    Rust toolchain with Cargo and Git are required. The Quick Start shows cloning the repository and starting the service with a config.yaml, or using Docker to run the prebuilt image. Optional: Inspector for debugging via Node.js (npx @modelcontextprotocol/inspector).

    Details
    Last Updated1/1/2026
    SourceGitHub

    Compare Alternatives

    Similar MCP Tools

    9 related tools
    Playwright MCP

    Playwright MCP

    Playwright MCP server. A Model Context Protocol (MCP) server that provides browser automation capabilities using Playwright. This server enables large language models (LLMs) to interact with web pages through structured accessibility snapshots, bypassing the need for screenshots or visually-tuned models. The server is designed to be fast, lightweight, and deterministic, offering LLM-friendly tooling and a rich set of browser automation capabilities via MCP tools. It supports standalone operation, containerized deployments, and integration with a variety of MCP clients (Claude Desktop, VS Code, Copilot, Cursor, Goose, Windsurf, and others).

    Sequential Thinking MCP Server

    Sequential Thinking MCP Server

    Sequential Thinking MCP Server provides a dedicated MCP tool that guides problem-solving through a structured, step-by-step thinking process. It supports dynamic adjustment of the number of thoughts and allows revision and branching within a controlled workflow, making it ideal for complex analysis and solution hypothesis development. This server is designed to register a single tool, sequential_thinking, and is integrated with common MCP deployment methods (NPX, Docker) as well as editor integrations like Claude Desktop and VS Code for quick setup. The documentation provides exact configuration snippets, usage patterns, and building instructions to help you deploy and use the MCP server effectively, including Codex CLI, NPX, and Docker installation examples.

    N8N MCP Server

    N8N MCP Server

    An MCP (Model Context Protocol) server designed to integrate Claude Desktop, Claude Code, Windsurf, and Cursor with n8n workflows. This MCP enables users to build, test, and orchestrate complex workflows by exposing a set of tools that bridge Claude’s capabilities with n8n’s automation platform. The project emphasizes robust trigger handling, multi-tenant readiness, and progressive documentation to help developers understand how tools map to real-world workflow tasks. It also outlines future tooling integration points (such as getNodeEssentials and getNodeInfo) to further enhance node-structure awareness within MCP-powered automations.

    Hugging Face MCP Server

    Hugging Face MCP Server

    Hugging Face Official MCP Server connects your large language models (LLMs) to the Hugging Face Hub and thousands of Gradio AI Applications, enabling seamless MCP (Model Context Protocol) integration across multiple transports. It supports STDIO, SSE (to be deprecated but still commonly deployed), StreamableHTTP, and StreamableHTTPJson, with the Web Application allowing dynamic tool management and status updates. This MCP server is designed to be run locally or in Docker, and it provides integrations with Claude Desktop, Claude Code, Gemini CLI (and its extension), VSCode, and Cursor, making it easy to configure and manage MCP-enabled tools and endpoints. Tools such as hf_doc_search and hf_doc_fetch can be enabled to enhance document discovery, and an optional Authenticate tool can be included to handle OAuth challenges when called.

    Shadcn UI MCP Server v4

    Shadcn UI MCP Server v4

    Shadcn UI v4 MCP Server is an advanced MCP (Model Context Protocol) server designed to give AI assistants comprehensive access to shadcn/ui v4 components, blocks, demos, and metadata. It enables multi-framework support (React, Svelte, Vue, and React Native) with fast, cache-friendly access to component source code, demos, and directory structures, empowering AI-driven development workflows. The project emphasizes production-readiness with Docker Compose, SSE transport for multi-client deployments, and smart caching to optimize GitHub API usage while providing rich metadata and usage patterns for rapid prototyping and learning across frameworks.

    Figma MCP server

    Figma MCP server

    The Figma MCP server enables design context delivery from Figma files to AI agents and code editors, empowering teams to generate code directly from design selections. It supports both a remote hosted server and a locally hosted desktop server, allowing seamless integration with popular editors through Code Connect and a suite of tools that extract design context, metadata, variables, and more. This guide covers enabling the MCP server, configuring clients (VS Code, Cursor, Claude Code, and others), and using a curated set of MCP tools to fetch structured design data for faster, more accurate code generation. It also explains best practices, prompts, and integration workflows that help teams align generated output with their design systems. The documentation includes concrete JSON examples for configuring servers in editors like VS Code and Cursor, as well as command examples for Claude Code integration and plugin installation.

    MarkItDown MCP

    MarkItDown MCP

    MarkItDown-MCP is a lightweight MCP (Model Context Protocol) server provided as the markitdown-mcp package. It exposes a STDIO, Streamable HTTP, and SSE MCP server designed for calling MarkItDown to convert content to Markdown. The package focuses on simplicity and accessibility, enabling you to run the MCP server locally via a simple CLI, or in Docker for containerized workflows, with integration options for Claude Desktop. The core capability is exposed through a single tool, convert_to_markdown(uri), which accepts a URI in http:, https:, file:, or data: schemes to fetch content and convert it to Markdown. This MCP server is easy to install with pip and can be used in various transport modes, including STDIO and HTTP/SSE, making it a flexible choice for automations and integrations.

    Chrome MCP Server

    Chrome MCP Server

    Chrome MCP Server is a Chrome extension-based Model Context Protocol (MCP) server that exposes your Chrome browser functionality to AI assistants like Claude, enabling complex browser automation, content analysis, and semantic search. It leverages your existing Chrome environment, including login states and configurations, to allow large language models and chatbots to control the browser natively without needing to launch a separate automation process. The project emphasizes privacy by remaining fully local and offers capabilities such as cross-tab context, streamable HTTP communication, and a built-in vector database for semantic search and content analysis. As an early-stage project, it includes a growing set of tools for browser control, inspection, and automation, with ongoing development to broaden compatibility and features.

    MCP server for Appwrite docs

    MCP server for Appwrite docs

    The MCP server for Appwrite docs enables LLMs and code-generation tools to interact with comprehensive Appwrite documentation. It empowers AI assistants to access up-to-date API references, SDK guides, and implementation examples, facilitating intelligent code generation, troubleshooting, and best-practice guidance directly from the official docs. This MCP brings real-time context, semantic search, and seamless integration with popular editors and IDEs to accelerate development workflows around Appwrite's APIs and SDKs.