Codex

Give Codex the ability to create cloud sandboxes, run code, deploy applications, and expose services to the internet.

Setup

1. Install Conway Terminal

curl -fsSL https://conway.tech/terminal.sh | sh

Or install manually:

npm install -g conway-terminal

2. Configure Codex

Add Conway as an MCP server in your Codex configuration:

codex --mcp-config conway-mcp.json

Create conway-mcp.json:

{
  "mcpServers": {
    "conway": {
      "command": "conway-terminal",
      "env": {
        "CONWAY_API_KEY": "cnwy_k_..."
      }
    }
  }
}

Your API key is in ~/.conway/config.json after running the setup script.

Verify Integration

Ask Codex to list your sandboxes:

> list my cloud sandboxes

Codex should invoke the sandbox_list tool.

Example Workflows

Run Tests in Isolation

> create a sandbox, copy my project, install dependencies, and run the test suite

Deploy a Prototype

> create a sandbox with 2 vCPUs, set up a Flask API, and expose port 8000

Multi-Model Inference

> use Conway Compute to compare responses from GPT-5.2 and Claude Sonnet 4.5 for this prompt

Tips

• Codex can use Conway's sandboxes as isolated execution environments for running and testing code
• Chain sandbox operations with inference calls for agentic workflows
• Use PTY sessions for interactive debugging

Troubleshooting

Tools Not Found

1. Verify conway-terminal is installed: which conway-terminal
2. Check the MCP config path is correct
3. Ensure the API key is valid

Connection Issues

• Test connectivity: curl https://api.conway.tech/health
• Verify your API key: check ~/.conway/config.json