analyze
Analyze recent runs - per-model latency, tokens, and verdict agreement - and suggest model/reasoning/fanout tuning. Advisory, read-only.
Log GPT (Codex) in on this machine with a ChatGPT device login - shows a link and a one-time code. For remote or headless sessions (Claude Code on the web). No question is sent to GPT.
> /plugin marketplace add antonbabenko/deliberation > /plugin install deliberation@antonbabenko-deliberation
How it fires
How this command gets triggered: by you, by Claude, or both.
/codex-loginContext preview
What this command does when you run it.
Log GPT (Codex) in on this machine with a ChatGPT device login - shows a link and a one-time code. For remote or headless sessions (Claude Code on the web). No question is sent to GPT.
name: codex-login description: Log GPT (Codex) in on this machine with a ChatGPT device login - shows a link and a one-time code. For remote or headless sessions (Claude Code on the web). No question is sent to GPT. allowed-tools: mcp__deliberation__codex-login timeout: 60000
Starts, or joins, the ChatGPT device login for GPT on this machine and shows its link and one-time code. Nothing is asked of GPT. Use it when GPT has no login here, typically in a Claude Code web session.
1. Call the tool, with no arguments:
mcp__deliberation__codex-login({})2. Show the result's `message` to the user **as-is**: never paraphrased, never truncated, and never reflowed - the link and the code each stand on their own line so they are one copy. By `status`:
link, sign in to ChatGPT, and enter the code; GPT answers from their next GPT call.
dialog. The message also says what to do if an existing credential turns out to be spent (a copied `auth.json`): the next GPT call replaces it, or `codex logout` first.
same login), up to 3 times, before showing anything; only if it is still `starting`, show the message and ask the user to run `/deliberation:codex-login` again shortly.
login, and the next call offers a new code.
disabled in config, a login that ended). 3. Stop. Do not run shell commands, do not read or check `~/.codex/auth.json`, and do not call any other deliberation tool in the same turn.
Present them plainly; do not summarize them away.
the same login and shows the same code until it expires (15 minutes).
`auth.json` from another machine: a ChatGPT refresh token works once, so a shared copy breaks on the first refresh.
Get a second opinion in Claude Code from GPT, Gemini, and Grok - plus 400+ more models through OpenRouter, including Qwen, Kimi, and DeepSeek.
Repo: antonbabenko/deliberation
Analyze recent runs - per-model latency, tokens, and verdict agreement - and suggest model/reasoning/fanout tuning. Advisory, read-only.
Ask GPT, Gemini, Grok, and any configured OpenRouter models in parallel for independent second opinions, then synthesize and compare. Zero cross-contamination.
Get Gemini second opinion on a question or current work. Single-shot, advisory, no contamination. Model pinned per call.
Get GPT (Codex) second opinion on a question or current work. Single-shot, advisory, no contamination.
Get Grok (xAI) second opinion on a question or current work. Single-shot, advisory, no contamination.
Ask a single configured OpenRouter model for a second opinion. Advisory only. Single-shot or multi-turn.