Install and connect

Supported platforms

Prebuilt packages are available for macOS and Linux on ARM64 and x86-64, and for Windows x64. Sandboxed Windows sessions require Windows 11 with Microsoft Execution Containers (MXC).

If Console fails to start on Linux, see Linux host compatibility for missing libraries or sandbox restrictions.

R, Python, and SQL

To let your agent use R, install it separately, for example with rig and rig add release. Python and SQL work without R. The first session may download Python and packages.

To verify the selected runtimes and prepare caches before opening a client, follow Check your installation.

To install the development version, follow Build from source.

Codex and Claude Code

Run the command for your client:

uvx mcp-console configure-client codex
uvx mcp-console configure-client claude

This registers Console globally, so it is available across projects, and loads its tools up front. It also adds Console’s send tool to the client’s allowlist so calls do not need individual approval, subject to explicit client restrictions. Setup starts runtime checks and cache warming in the background, printing a log path for the results. Use --no-warmup to skip that step. Restart the client after setup. Your client launches Console on the same host, usually in the workspace where the agent is working.

To register Console for a specific project, run the command from the project directory and add --directory .. See client configuration for project registration, version pinning, and launch options.

Other MCP clients

Configure a stdio server with command uvx and arguments ["mcp-console", "serve"]. Clients using an mcpServers mapping can use:

{
  "mcpServers": {
    "console": {
      "command": "uvx",
      "args": ["mcp-console", "serve"]
    }
  }
}

Console runs on the same host as the client, usually in the agent’s workspace. For remote work, run both on the remote host; Console does not manage remote connections or containers.

R and Python clients

To use Console from R or Python, follow Use Console from R or Use Console from Python. These clients use the same workbench and settings.

Build a data assistant shows how to change the tool descriptions, exposed languages and parameters, and prepared environment. Tool configuration documents the exact settings.

After connecting, Sessions and languages explains how your agent can work with Console.