Metadata-Version: 2.4
Name: llmx-worker
Version: 0.1.0
Classifier: Environment :: Console
Classifier: Operating System :: MacOS
Classifier: Operating System :: POSIX :: Linux
Classifier: Programming Language :: Rust
Summary: Share your Claude Code or Codex capacity on llmx.tw and get paid per token served.
Requires-Python: >=3.8
Description-Content-Type: text/markdown; charset=UTF-8; variant=GFM
Project-URL: Homepage, https://llmx.tw

# llmx-worker

The supplier side of [llmx](https://llmx.tw): share your Claude Code or Codex
capacity with other llmx users and get paid for every token you serve.

## Install

    uv tool install llmx-worker        # or: pipx install llmx-worker

Prebuilt for macOS (Apple silicon and Intel) and Linux (x86_64 and arm64).
Upgrade with `uv tool upgrade llmx-worker`.

## Run

Sign in at https://llmx.tw/dashboard, open **Share capacity**, pick a
provider and create a worker token. Then:

    # Claude: prompts run here under your own Claude Code login
    claude auth login
    llmx-worker --provider claude --token wkr_YOUR_TOKEN

    # Codex: uses the login from `codex login`
    codex login
    llmx-worker --provider codex --token wkr_YOUR_TOKEN

The worker connects to `wss://llmx.tw/ws/worker` by default. Stop it with
Ctrl-C: it finishes what it is serving, then exits. `llmx-worker --help`
lists every option.

## What runs on your machine

- **Claude:** each request runs through your local `claude` CLI with all of
  its built-in tools turned off, in an empty temporary folder. It cannot read
  your files or run commands; tool calls go back to the consumer's own
  machine. Your Claude login never leaves your machine.
- **Codex:** your Codex login is uploaded to llmx once, encrypted. The worker
  only relays encrypted traffic between llmx and OpenAI, and connects only to
  OpenAI's hosts.

Sharing a subscription may break your provider's terms of service; you
accept that risk by running a worker.

