crit for VS Code

io.github.crit-labs/crit

Build, evaluate and serve a bespoke LLM inference engine for your own GPU with crit's commands.

client:VS Code transport:stdio runtime:pypi

Install crit in VS Code

.vscode/mcp.json

{
  "servers": {
    "crit": {
      "type": "stdio",
      "command": "uvx",
      "args": [
        "crit"
      ]
    }
  }
}

Requires agent mode. MCP: Open User Configuration puts the same block in every workspace. VS Code docs