A discord frontend for Codex CLI and Claude Code
  • Python 86.9%
  • Shell 5.9%
  • Dockerfile 3.6%
  • Slim 3.6%
Find a file
2026-02-26 02:15:26 +01:00
docs vibe-modernize (v2 basically) 2026-02-24 10:57:14 +01:00
.example.env change default model 2026-02-24 16:51:43 +01:00
.gitignore ignore disabled env 2026-02-26 02:01:12 +01:00
.gitmodules remove dead openai-websearch-mcp submodule 2025-08-22 22:37:27 +02:00
bot.py fix bot attaching files 2026-02-26 02:15:26 +01:00
codex.example.env vibe-modernize (v2 basically) 2026-02-24 10:57:14 +01:00
Dockerfile remove my web search tool, codex has builtin support for that 2026-02-24 16:05:04 +01:00
Dockerfile.slim remove my web search tool, codex has builtin support for that 2026-02-24 16:05:04 +01:00
entrypoint.sh misc changes I forgot to commit for weeks (ooops) and claude code support (new) 2025-06-23 00:40:25 +02:00
launch_services.sh vibe-modernize (v2 basically) 2026-02-24 10:57:14 +01:00
LICENSE fix license and git clone link in SETUP.md 2025-06-02 11:56:09 +02:00
README.md vibe-modernize (v2 basically) 2026-02-24 10:57:14 +01:00
requirements.txt feat: when messages exceed input size limit, split into multiple messages; feat: file uploads 2025-06-28 11:10:35 +02:00

CodexMaster

... 'cause my 16yo ass can't afford $200/month😐

What is this?

Simply put, it's a Discord frontend for Codex CLI. It does a bit more than that though, because you can spawn and manage multiple agents at once, so it's actually more like the cloud-based Codex in ChatGPT, I think.

Why would you even need this?

Well, I don't know about you, but I don't always have immediate access to a terminal where I can run all of these tools in a properly isolated environment.

Some more info

This version uses the standard codex CLI directly (no codex-headless fork and no Claude/Gemini integrations). If you run agents in Docker, the image installs @openai/codex and mounts your host ~/.codex into the container so Codex auth/config/session data are shared.

How to use

Roughly, you have to

  1. Create a discord bot. This is supposed to be self hosted and access restricted (you have to whitelist users in .env). As you may be able from the code, this system is not built to be scaled.
  2. Populate .env file with bot token, whitelist, any API keys (openai, anthropic, ...), and more. See .example.env.
  3. Run directly with python bot.py, create a systemd service or a docker container, whatever. Recommended option for deployment on linux is a systemd service that runs as a non-root user.

Execution mode is now chosen per agent at /spawn time (docker or host). Use ALLOW_DOCKER_EXECUTION, ALLOW_HOST_EXECUTION, and DEFAULT_EXECUTION_MODE in .env to control which modes are available and which one is the default.

Agent notification detail is also configurable per spawn via verbosity (answers vs verbose). Set DEFAULT_AGENT_VERBOSITY in .env to choose the default.

codex.env is optional. It is only a convenience file for passing API keys / Codex env flags into agent runs (especially Docker mode or host mode with leak_env=false).

Requirements

  • python (required) >= 3.9
  • node.js (optional, host mode only) >= 22
  • rootless docker (recommended) >= 21: If you want to run Codex in a docker container to minimize potential damage. I recommend rootless docker.
  • linux (recommended): CodexMaster was developed on linux and is only tested there. It is meant to be hosted as a discord bot, after all.

More details

Setup: See SETUP.md.

Commands: See COMMANDS.md.

License

MIT License