coding.sh proxy

coding.sh proxy [action] [options]

  Run a local Anthropic/OpenAI API proxy that logs every model call

Commands
  status        Show whether a proxy is running, and where (exit 1 when not)
  stop          Stop the background proxy
  env           Print the ANTHROPIC_BASE_URL / OPENAI_BASE_URL exports for the proxy

Options
  -p, --port <PORT>               Port to listen on, 0 = any free port ([proxy] port) (default: 4320)
      --host <HOST>               Interface to bind ([proxy] host) (default: 127.0.0.1)
      --anthropic-upstream <URL>  Where Anthropic calls go, e.g. a gateway ([proxy] anthropic_upstream) (default: https://api.anthropic.com)
      --openai-upstream <URL>     Where OpenAI calls go ([proxy] openai_upstream) (default: https://api.openai.com)
      --background                Start detached, print the URL once it answers (log: ~/.coding.sh/proxy.log)
  -q, --quiet                     No line per call on stderr

Examples
  coding.sh proxy
  coding.sh proxy --background
  coding.sh init --proxy

Notes
  Forwards /v1/messages (and count_tokens), /v1/chat/completions, /v1/responses and every other path to the real API with the client's own auth headers. Auth headers are never stored.
  Streams pass through as they arrive. A failed store write never fails the request.
  Calls from Claude Code and Codex land in the session their hooks capture; other clients get a proxy-<hash> session per conversation (or send x-coding-sh-session: <id>).
  Prints the environment to stdout; one line per call goes to stderr.

coding.sh proxy status

  Show whether a proxy is running, and where (exit 1 when not)

Examples
  coding.sh proxy status

coding.sh proxy stop

  Stop the background proxy

Examples
  coding.sh proxy stop

coding.sh proxy env

  Print the ANTHROPIC_BASE_URL / OPENAI_BASE_URL exports for the proxy

Examples
  eval "$(coding.sh proxy env)"