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

MCP Connect

Community
Gateway / proxy
Lightweight bridge that exposes local MCP servers as HTTP APIs

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
Dev Tools #355 of 438
Stars
★ 240 Popular
Transport
stdio (local process)
Runtime
Node.js 22+ · Python
Credentials
API key / credential required
License
MIT
Last commit
⚠
40FMRS · D

MCP Connect provides a practical bridge solution for converting local MCP servers into HTTP-callable services. It addresses the limitation of cloud AI services that cannot directly use stdio protocol, supports multiple invocation modes, and enhances security. Suitable for developers needing remote integration of MCP tools, but care must be taken with security configurations. Open source under MIT license.

Strongest · Documentation 12/20 Weakest · Security and permissions 6/20

Reliability
8/20
Security and permissions
6/20
Maintenance
6/20
Documentation
12/20
Setup experience
8/20
Why each score
Reliability 8/20
Based on the README, the project claims to bridge stdio MCP servers to HTTP, with streaming HTTP and classic request/response modes. Source files include http-server.ts and mcp-client-manager.ts, but no tests or CI configuration are provided, so we cannot statically verify that key paths run reliably. Deductions for missing test evidence and error handling details.
Security and permissions 6/20
Bearer token authentication is implemented, with ACCESS_TOKEN required at startup, which meets basic security. However, there is no visible restriction on token storage, CORS details, or protection against SSRF. The bridge can execute arbitrary commands via tool calls, and there is no confirmation mechanism, lowering the security score.
Maintenance 6/20
The repository has 238 stars and only 1 open issue, but recent commits and release info are not provided. MIT license is present, but maintenance activity and versioning are not visible. Deductions for lack of visible release history and dependency update policy.
Documentation 12/20
The README provides clear architecture diagrams, feature tables, and quick start steps, including environment variables and API references. However, it lacks in-depth troubleshooting guides, detailed tool parameter explanations, and cost disclosures. No test evidence for examples is provided, so the documentation score is moderate.
Setup experience 8/20
Installation steps are clear: clone, npm install, configure environment variables and mcp-servers.json. For streaming HTTP mode, manual configuration file editing is required, but examples are provided. Lacks client configuration examples and requires Node.js version, adding slight complexity.

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

  • Developers needing to integrate local MCP servers with cloud AI services
  • Those needing to bridge existing stdio MCP servers over HTTP
  • Users requiring session management and secure authentication

Not for

  • Scenarios requiring fully native MCP protocol without extra bridging
  • Pure local MCP use without remote access
  • Users seeking minimal configuration

Required permissions

  • Requires setting ACCESS_TOKEN environment variable
  • May need to configure ALLOWED_ORIGINS to restrict cross-origin
  • If using Ngrok, provide NGROK_AUTH_TOKEN
  • If using E2B, provide E2B_API_KEY

Risks and side effects

  • Exposing local services to the public internet carries security risks
  • Bridging may introduce latency
  • Misconfiguration may prevent the server from starting

Setup

Before you start

Runtime：Node.js 22+ · Python

ACCESS_TOKEN requiredsecret Required Bearer auth token for the local service; set it yourself in .env — startup fails if empty.
NGROK_AUTH_TOKEN optionalsecret Obtain from https://dashboard.ngrok.com/get-started/your-authtoken to enable the public tunnel.
E2B_API_KEY optionalsecret API key obtained after signing up at https://e2b.dev, used for E2B sandbox deployment.
E2B_MCP_AUTH_TOKEN optionalsecret Self-chosen Bearer token for the bridge in E2B deployments.
Other optional settings (2)
ALLOWED_ORIGINS optional Comma-separated list of origins to restrict cross-origin requests in production; set it yourself.
LOG_LEVEL optional Log verbosity level (DEBUG/INFO/WARN); set as needed.
  1. Clone the repo: git clone https://github.com/EvalsOne/MCP-connect.git && cd mcp-connect
  2. Install dependencies: npm install
  3. Copy .env.example to .env and set ACCESS_TOKEN (required).
  4. (Optional) Edit mcp-servers.json to configure MCP servers for Streamable HTTP mode.
  5. Build and start: npm run build && npm start.
  6. Check http://localhost:3000/health to verify status.

Check that it works

After starting the server, visit http://localhost:3000/health; a response of {"status": "ok"} confirms it is running. Then connect an MCP client to http://localhost:3000/mcp/<serverId> or send a tools/list request to /bridge to confirm tools are visible.

Troubleshooting

  1. Verify ACCESS_TOKEN is set, otherwise startup fails
  2. Ensure mcp-servers.json is configured, otherwise Streamable HTTP endpoints are unavailable
  3. Check /health endpoint for server status
  4. Inspect combined.log or error.log for details

Things to try

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

  • Fetch the content of https://example.com via the /mcp/fetch endpoint
  • List the tools exposed by the configured MCP server using tools/list
  • Call the fetch tool of mcp-server-fetch through the /bridge endpoint to read a webpage
  • List the prompt templates supported by this MCP server with prompts/list

Use cases

Allow cloud AI tools to access local MCP servers
Expose MCP servers as HTTP endpoints for remote clients
Provide public access to local MCP via Ngrok
Quickly deploy MCP servers in E2B cloud sandbox

Overview

MCP Connect is an HTTP gateway that lets you call local MCP servers (that speak stdio) through Streamable HTTP or a classic request/response bridge. It supports session management, Bearer token authentication, CORS allowlist, built-in Ngrok tunnel, and one-click deploy to E2B cloud sandbox. Features dual modes: Streamable HTTP mode with unique routes per MCP server, and a non-standard /bridge endpoint. Suitable for integrating local MCP tools into remote cloud AI clients.

Similar servers

Context7 80 · B

Upstash's official server providing up-to-date third-party library docs for AI coding assistants

★ 62.9k · Tools 2 Compare with this →
Compare with other MCP servers

Source revision 11c111c0fc54 Data synced 2026-10-12 Read the FMRS scoring method