M
MCP Directory CURATED DIRECTORY
911 MCP servers indexed
中文
← Back to directory
M

MCPHub

Community
Gateway / proxy
Unified hub for centrally managing multiple MCP servers

This gateway aggregates or proxies other MCP servers; it only becomes useful once you connect real servers behind it.

How to install ↓ Compare GitHub source repository ↗
Category
Other #195 of 230
Stars
★ 2.5k Very popular
Transport
Streamable HTTP
Runtime
Docker · Node.js
Credentials
API key / credential required
License
Apache-2.0
Last commit
42FMRS · D

MCPHub is a feature-rich MCP gateway offering powerful centralized management and routing capabilities. It suits teams that need to manage multiple MCP servers but requires attention to security configuration and resource consumption.

Strongest · Maintenance 10/20 Weakest · Reliability 7/20

Reliability
7/20
Security and permissions
8/20
Maintenance
10/20
Documentation
9/20
Setup experience
8/20
Why each score
Reliability 7/20
Evidence: The repo includes complete source code, package.json, pnpm workspace, CI workflows, and test directories, but static review does not verify actual runtime. Core features like server management, routing, and WebSocket/SSE endpoints have implementations, but error handling and edge-case test coverage are not clearly evidenced, and no reproducible handshake logs are provided. Deductions: No execution, missing clear test coverage for key paths and diagnostic details, so reliability below 12.
Security and permissions 8/20
Evidence: Code uses JWT, bcrypt, OAuth libraries, supports bearer token authentication, and docs explicitly state authentication is required by default. However, static review finds that credential storage and management details (e.g., session keys, database credentials) are not fully disclosed, and the granularity of API key permissions (e.g., all/read-only) and confirmation mechanisms for dangerous operations are not fully verified. No signs of malware or credential theft. Deductions: Incomplete evidence on least-privilege model and explicit confirmation for dangerous operations, so not full marks.
Maintenance 10/20
Evidence: Repo is active with recent commits and version tags, issues are responded to (34 open issues), license is Apache-2.0, and there is some evidence of dependency updates via GitHub Actions. Deductions: No clear security response channel and detailed release policy documentation, so not full marks.
Documentation 9/20
Evidence: README is detailed, covering quick start, Docker deployment, configuration guide, and CLI usage, with external docs link. It covers main features and authentication. Deductions: Full API reference (error codes, rate limiting), troubleshooting, and cost information are missing; some advanced configurations are not detailed.
Setup experience 8/20
Evidence: Docker images and quick start commands provided, configuration via mcp_settings.json is clear, client connection URL examples given. Local development commands provided. Deductions: No automated configuration script or full client configuration examples for mainstream clients (e.g., Claude Desktop), steps are many, dependencies on Docker and Node/TypeScript, and no verification of execution.

Static review · not runListed 2026-08-07

Read the FMRS scoring method →

Fit and risk

What it can accessRuns commands or codeUses the network

Best for

  • Development teams needing centralized management of multiple MCP servers
  • Organizations exposing various MCP servers to AI clients via a unified API
  • Production deployments requiring high availability and scalability
  • Solo developers wanting to quickly experiment with various MCP servers

Not for

  • Simple scenarios requiring only a single MCP server
  • Users with strict customization needs for a specific MCP server
  • Environments where running without authentication is unacceptable (high security requirements)

Required permissions

  • Access to MCP server configuration files and local data directory
  • Ability to execute commands like npx and uvx to start MCP servers
  • Network access to communicate with MCP servers
  • Optional: PostgreSQL database connection for database mode

Risks and side effects

  • MCP endpoints require authentication by default; unauthenticated access may expose servers
  • Hot-swappable configuration could lead to configuration errors causing service unavailability
  • Running many MCP servers may consume significant system resources
  • Admin password and API keys need careful management

Setup

Before you start

Runtime：Docker · Node.js

ADMIN_PASSWORD optionalsecret Sets the initial password for the dashboard admin account; if unset, a random password is generated and printed to server logs.
Other optional settings (1)
DISABLE_WEB optional Set to true to skip the bundled dashboard UI and run only the backend/API and MCP endpoints (headless mode).
  1. Create an mcp_settings.json file with your MCP server configurations.
  2. Run with Docker: docker run -p 3000:3000 -v ./mcp_settings.json:/app/mcp_settings.json -v ./data:/app/data samanhappy/mcphub.
  3. Access http://localhost:3000, log in with username 'admin'; on first launch, a random password is generated and printed to logs, or you can pre-set it via environment variable ADMIN_PASSWORD.
  4. Connect AI clients to endpoints: http://localhost:3000/mcp (all servers), /mcp/{group} (group), /mcp/{server} (specific server), /mcp/$smart (smart routing).

Check that it works

After starting the container, open http://localhost:3000 and log in as admin to see the dashboard; or configure the http://localhost:3000/mcp endpoint in an AI client — its tool list showing the connected servers' tools confirms the connection.

Troubleshooting

  1. Common issues: 1. Ensure mcp_settings.json is well-formed and server commands are executable. 2. Check that Docker volumes for config and data are correctly mounted. 3. Confirm port 3000 is not in use. 4. If unable to log in, check generated password in logs or set ADMIN_PASSWORD. 5. If MCP endpoints are inaccessible, review authentication settings.

Things to try

Once connected, you can ask your AI assistant things like:

  • List all MCP servers currently connected to the hub
  • Use $smart routing to find a tool that can fetch a web page and grab https://example.com
  • Add the time server to a group called daily
  • Call the fetch_url tool with url=https://example.com

Use cases

Manage and access multiple MCP servers under one unified endpoint
Route per group or individual server, controlling visibility of tools/prompts/resources
Leverage smart routing for AI-powered tool discovery
Secure authentication via OAuth 2.0 and social login
Rapid deployment in Docker environments

Supported clients

Claude Desktop
Cursor

Listed from the project's documentation, not tested by this site.

Overview

MCPHub is a unified hub for centrally managing and dynamically orchestrating multiple MCP servers/APIs into separate endpoints with flexible routing strategies. It offers a centralized dashboard, flexible routing, granular control over tool/prompt/resource visibility per group, AI-powered tool discovery via vector semantic search, MCP Apps proxy, tool result compression, hot-swappable configuration, OAuth 2.0 support, social login, database mode with PostgreSQL, and Docker-ready deployment.

Similar servers

BioMCP 75 · B

One binary. One grammar. Evidence from the biomedical sources you already trust.

★ 654 · Tools 11 Compare with this →
Compare with other MCP servers

Source revision 0ee8100ec291 Data synced 2026-10-12 Read the FMRS scoring method