greed-compute-mcp

MCP server that gives LLMs stateful Python execution. Connect it to Claude, GPT, or any MCP-compatible agent and they can create sessions, run code, checkpoint state, fork workers, and run parallel MapReduce — all through natural tool calls.

Setup

npm install -g greed-compute-mcp

Set your API key:

export GREED_API_KEY=gc_your_key_here

Usage with Claude Desktop

Add to your claude_desktop_config.json:

{
  "mcpServers": {
    "greed-compute": {
      "command": "greed-compute-mcp",
      "env": {
        "GREED_API_KEY": "gc_your_key_here"
      }
    }
  }
}

Usage with Claude Code

claude mcp add greed-compute -- greed-compute-mcp

Set the env var in your shell or .env.

Available Tools

| Tool | What it does | |------|-------------| | create_session | Spin up a Python interpreter (blank, data-science, ML, or scraping template) | | execute_code | Run Python in a session. State persists between calls | | session_status | Check TTL and activity | | terminate_session | Kill a session | | install_packages | pip install into a running session | | create_checkpoint | Snapshot all interpreter state | | restore_checkpoint | Load a checkpoint into a session | | list_checkpoints | See all saved checkpoints | | delete_checkpoint | Free checkpoint storage | | create_swarm | Parallel MapReduce across N workers | | get_swarm | Poll swarm progress | | create_workspace | Persistent shared environment | | execute_in_workspace | Run code in shared workspace | | list_workspaces | See your workspaces | | upload_file | Push files into a session | | download_file | Pull files out of a session | | get_usage | Check plan limits and usage |

Environment Variables

| Variable | Required | Default | |----------|----------|---------| | GREED_API_KEY | Yes | — | | GREED_COMPUTE_URL | No | https://compute.deep-ml.com |

License

MIT