Mcpd-Proxy: Centralized Tool Access for AI Agents in VS Code, Cursor, and Beyond (blog.mozilla.ai)

Mozilla.ai announced mcpd-proxy, a lightweight MCP (Model Context Protocol) server that provides a single, IDE-friendly entry point to centrally managed mcpd tool servers. The build grew out of enterprise feedback (notably from Microsoft’s AI consulting teams) about the “last mile” problem: platform teams can provision and version MCP servers centrally, but developers still had to add/maintain many per-server IDE configs. mcpd-proxy solves that by connecting to your central mcpd instance and exposing all managed servers and tools through one endpoint developers register once (e.g., an mcp.json entry using npx @mozilla-ai/mcpd-proxy with MCPD_ADDR and MCPD_API_KEY). Technically, mcpd is a declarative manager for MCP servers (sidecar pattern) that abstracts tool configuration, secrets, and versioning. mcpd-proxy acts as another MCP server that federates those managed servers, namespacing tools (github__list_pull_requests, slack__slack_post_message) so agents in VS Code or Cursor can call cross-server tools transparently. The pattern preserves central control—available tools, server versions, access policies, rate limits and health monitoring—while delivering zero-config onboarding for developers. This reduces config drift, local processes, and version mismatch risk. It’s an early release intended for platform teams managing agent toolchains; the team is soliciting feedback on where it helps and breaks.