dockashell

DockaShell provides each MCP client with isolated, persistent Docker containers that act as their working environments. Each agent receives its own container with a persistent volume for files, databases, and context, plus shell access and full audit trails of commands and changes. Agents can run standard POSIX tools directly inside these containers (bash, git, npm, Python, etc.), create and modify tools, code, and data, and evolve their workflows over time without affecting the host system. Use the MCP client to connect to the dockashell MCP server and start a dedicated project workspace; the server will spin up a container, mount persistent storage, and begin logging all activity for analysis and self-improvement. The included tools and documentation cover common workflows like data analysis, web development, and research task automation, enabling agents to build end-to-end pipelines inside their own isolated environments.

Prerequisites:

• Node.js 20+ and npm installed on your host
• Docker daemon running and accessible
• Access to the internet for npm package installation

Step 1: Install the DockaShell CLI globally

npm install -g dockashell

Step 2: Build and set up a workspace (example project name used here)

dockashell build
dockashell create my-project

Step 3: Start the DockaShell server for the workspace

dockashell start my-project

Step 4: Add the MCP client configuration to your agent or MCP client

{
  "mcpServers": {
    "dockashell": {
      "command": "dockashell",
      "args": ["serve"]
    }
  }
}

Note: You can customize the configuration, including environment variables and paths, via the docs at docs/dockashell-configuration.md and docs/mcp-tools.md.

Tips and common considerations:

• Ensure the Docker daemon is accessible to the DockaShell process (permissions or user groups may be required).
• The server provides a persistent workspace per agent; understand your storage quotas and backup strategies for long-running experiments.
• Audit trails are generated for each command and file change to aid analysis and reproducibility.
• Environment variables can be used to customize Node.js version requirements and Docker host configuration when running in different environments.
• If you encounter connectivity issues between the MCP client and the server, verify network access and that the command path (dockashell) is resolvable in the client environment.
• Review the documentation sections: CLI usage, Configuration, and MCP tools for deeper workflow examples and advanced settings.