AI NewsDev toolsAnnouncement

Moxie launched a Mac menu-bar app that lets Claude Code use ChatGPT, Gemini or a local model and swap accounts mid-task

Birju Vachhani launched Moxie on Product Hunt on 2 October, a Mac menu-bar app that routes Claude Code to any OpenAI-compatible provider and switches accounts mid-session when one hits its five-hour limit.

AI News

Editorial3 min read

LinkedInX
Moxie app screenshot showing Claude Max and Codex Plus accounts with five-hour and weekly usage limits side by side in a Mac menu bar

Image: Moxie

Why it mattersA developer whose Claude Code work gets interrupted every few hours by a usage limit now has a one-click way to continue the same session against a different plan or a local model.

A developer paying for Claude Max still gets kicked out of a long task every few hours, and starting a new session on a different account loses the context they just built up.

Birju Vachhani launched Moxie on Product Hunt on 2 October 2026 to answer that. It is a Mac menu-bar app that sits between Claude Code and the model behind it, points Claude Code at localhost:8060 through the ANTHROPIC_BASE_URL environment variable, and lets a developer swap the active provider mid-session. The listing had 77 upvotes within a day on Product Hunt, where it is ranked under AI Infrastructure Tools and LLM Developer Tools.

What a user can do with it

A Moxie user can plug in Claude plans, a ChatGPT plan, Gemini, DeepSeek, API keys for OpenRouter or any OpenAI-compatible endpoint, or a local model running under Ollama or Apple on-device, Vachhani writes on the site. If a developer is already signed in to Claude Code, Codex or Anthropic's Antigravity on the same Mac, Moxie can import those logins directly.

The switch itself is one click: a request already running finishes on the account that started it, and the next request goes to whichever provider the user picked. The five-hour and weekly usage windows for every connected account sit side by side in the menu bar, with a notification at 90% of a limit, another when it runs out and one more when it resets.

Local storage and no account

Vachhani says the app runs entirely on the Mac. API keys and login tokens live in the macOS Keychain, requests go straight from the Mac to the chosen provider, and there is no Moxie account. The only outbound traffic Vachhani lists is a licence or trial check and a daily signed-update ping. A 12-word BIP39 recovery phrase encrypts the local history and, Vachhani writes, is never sent off the machine.

The app also keeps a per-day heatmap of token use and a per-session log with request counts, token counts and data sent, which gives a user a running record of what each account cost on each task. A Playground view runs any connected model through a chat window, which Vachhani recommends for checking a new login before Claude Code depends on it.

Pricing

Moxie is free for three days with up to two connected accounts across Claude and OpenAI. A Standard licence is $39 one-off for two years and all providers, a Lifetime licence is $79 one-off with no renewal, both at a launch discount from $49 and $99. The licence does not include any AI usage: Moxie only routes Claude Code to accounts the user already has with each provider, and billing stays between the user and that provider.

It needs macOS 12 or later, and it sets up Claude Code for the user in both the CLI and the VS Code extension. If Claude Code rewrites its settings later, Moxie puts them back.

The user story Vachhani describes in the launch comment is a developer who spends most of a day in Claude Code and keeps hitting the five-hour cap halfway through a task. The alternatives today are a configuration file edit, a router install or a sign-out and sign-in cycle, each of which resets the agent's context. Moxie keeps that context and swaps the provider under it. Whether $39 is cheaper than losing two hours of agent time each day is a sum each developer runs for themselves.

Source

SourceMoxie

This item was written by an AI system from the linked source. Reveneau is responsible for what it publishes.

Share
LinkedInX