Mcpx FAQs

Question 1

What is Mcpx and what problem does it solve?

Accepted Answer

Mcpx is a gateway that consolidates multiple Model Context Protocol (MCP) servers into two simplified tools: 'search' and 'execute'. It significantly reduces AI agent context token consumption by preventing the agent from needing to load 100+ individual tool schemas, cutting tokens from potentially 84,000 to around 1,000.

Question 2

How does Mcpx achieve such a drastic reduction in context tokens?

Accepted Answer

Instead of providing numerous individual tool schemas to the agent, Mcpx exposes just two generic tools: `search` (to discover available tools and their parameters) and `execute` (to run agent-generated JavaScript code that calls the discovered tools). This 'Code Mode' approach, inspired by Cloudflare, saves vast amounts of context.

Question 3

Is executing agent-generated JavaScript code safe with Mcpx?

Accepted Answer

Yes, Mcpx executes agent-generated JavaScript code in a secure V8 isolate sandbox. This environment operates with deny-by-default permissions, meaning it has no access to the filesystem, network, or child processes. Tool calls are the only way code can interact with the outside world, and backend credentials are never exposed to the isolate.

Question 4

What deployment options are available for Mcpx?

Accepted Answer

Mcpx offers flexible self-hosted deployment options. For individual developers, it can run locally as a stdio subprocess. For teams, it can be deployed as an HTTP server using Docker, Kubernetes (Helm), or standalone, allowing multiple agents to connect to a single, centralized endpoint with token-based authentication.

Question 5

Which AI agents are compatible with Mcpx?

Accepted Answer

Mcpx supports any MCP-compatible agent that uses either stdio or HTTP (Streamable HTTP) for communication. Specific agents known to work include Claude Code, Cursor, Codex, Amp, and OpenCode, with straightforward configuration examples provided for each.

Mcpx

Mcpx

Key Features

Use Cases

Key Features

Use Cases