Deployment

This page covers the public deployment shapes for MUTX.

Local binaries

This is the simplest setup:

1. Run mutx-tui
2. Run mutx-mcp
3. Configure the agent to spawn mutx-mcp

Use this when the operator, the TUI, and the agent all live on the same host.

Split local TUI plus containerized MCP

This is useful when the agent expects an HTTP endpoint or when you want to keep the MCP runtime isolated.

Requirements:

• TUI running on the host
• shared socket directory mounted into the MCP container

Example:

docker run --rm \
  -p 8000:8000 \
  -v /tmp/mutx:/tmp/mutx \
  ghcr.io/mutx-net/mcp:latest

Fully containerized

You can containerize both components, but the TUI still needs:

• -it for interactive terminal access
• persistent config and data volumes
• a shared /tmp/mutx mount so the MCP container can reach the socket

Health checks

In HTTP mode, the MCP server exposes:

• /health for liveness
• /mcp for the actual MCP transport

Logging

mutx-tui can write structured logs to file and export them with:

mutx-tui logs export

Useful flags:

• --category
• --level
• --since
• --until
• --encrypt