Turning AutoDev into an MCP Service: Empowering AI Agents for End‑to‑End Coding

Within the broader discussion of Agentic Coding, the use of tools (Tool Use/Function calling) is a compelling topic. Completing a software development task often requires many external tools beyond the IDE, such as Git, Docker, Kubernetes, and Jenkins. Enabling AI to recognize and operate these tools is a key research direction.

From Progressive AI Agent Solutions to AutoDev as an MCP Service

In larger AI Agent scenarios, such as automated Computer Use, the IDE itself becomes a callable Agent tool. Current Agent Tool usage follows a progressive AI Agent model, because UI automation remains inefficient. Therefore, AutoDev focuses on IDE‑centered AI coding while also considering an Agent‑Tool‑centered approach.

Agent manipulates a browser to gather requirements.

Agent opens the IDE to write code.

Agent invokes DevOps tools to deploy code.

…

Since leveraging Agent Tools to call external utilities is a 2025 trend, AutoDev is offered as an MCP (Model Context Protocol) service, allowing any Agent Tool—including Cursor, Cline, or GitHub Copilot—to obtain high‑quality IDE context.

From MCP Ecosystem to AutoDev Commands

MCP, introduced by Anthropic for the Claude model, standardizes how LLM applications access external information, tools, and resources. Although AutoDev already provides extensive custom capabilities (e.g., Custom Agent), some tools remain unpredictable, and the MCP ecosystem is rapidly becoming a de‑facto standard.

AutoDev × MCP: Bidirectional Enablement

By combining MCP plugins (MCP Plugin) and the io.modelcontextprotocol ecosystem, AutoDev implements a two‑way enablement:

AutoDev acts as an MCP service callable by any Agent Tool.

AutoDev functions as an MCP client that can call any MCP service.

AutoDev as an MCP Service

Using JetBrains' MCP implementation, AutoDev can be configured as an MCP service via a simple JSON block. Below is a configuration example for the Cline plugin:

{
  "mcpServers": {
    "AutoDev": {
      "command": "npx",
      "args": ["-y", "@jetbrains/mcp-proxy"],
      "disabled": false,
      "autoApprove": []
    }
  }
}

The current version adds only database‑related capabilities; additional features will be added when suitable domestic MCP services become available.

AutoDev as an MCP Client

Similarly, AutoDev can act as a client by configuring MCP services on its Custom Agent page. The following JSON demonstrates the official MCP filesystem service:

{
  "mcpServers": {
    "filesystem": {
      "command": "npx",
      "args": ["-y", "@modelcontextprotocol/server-filesystem", "/Volumes/source/ai/auto-dev"]
    }
  }
}

Once configured, the MCP service’s tools become callable within AutoDev. For example, a DevIns command to list a directory returns JSON such as:

{
  "path": "/Volumes/source/ai/autocrud/docs/mcp"
}

With the powerful AutoDev DevIns command, any MCP service—or even an Agent—can be invoked directly from AutoDev.

Other Information

Life is short; I have AI.

Full documentation: https://ide.unitmesh.cc/mcp

Download the latest AutoDev (v2.0.0‑rc.2): https://github.com/unit-mesh/auto-dev/releases/tag/v2.0.0-rc.2