Magpie lets you run Claude Code on DeepSeek from the menu bar
Open-source tool Magpie just turned the messy world of coding agent configs into a clean menu bar app. It sets up a local gateway that lets you plug models like DeepSeek and Kimi into Claude Code or Codex, automatically fails over when you hit rate limits, and even scrubs API keys and passwords before prompts leave your machine.
Magpie is a free, open-source local proxy tool that brings model switching, account rotation, and gateway management for dozens of AI coding agents into your system menu bar. It lets you run Claude Code on Kimi or Codex on DeepSeek without manually touching config files or rewriting base URLs.
What it does
Most coding agents like Claude Code, Codex, and Gemini CLI keep their configurations in isolated files with distinct formats. Your paid subscription in one tool rarely works in another, and hitting a rate limit usually means a hard crash mid-task.
Magpie solves this by running a unified local gateway at 127.0.0.1:3425. It natively speaks and translates between OpenAI Chat, Anthropic Messages, and Gemini protocols, handling streaming, tool calls, and reasoning signatures seamlessly.
When you select a model from the menu bar, Magpie atomically updates the agent's native config file, preserving comments and formatting. Presets cover Anthropic, OpenAI, DeepSeek, Kimi, Zhipu GLM, and dozens of cloud providers. It can also detect existing local logins from Claude, ChatGPT, or Copilot and expose them as providers to other agents.
It also introduces routing groups with automatic failover. When an account runs out of tokens, Magpie quietly moves to the next account or a backup provider, shielding the agent from errors. Routing options range from smart pace tracking to exhausting plans that renew soonest.
To prevent leaking credentials, Magpie includes local prompt masking. When enabled, passwords, keys, and tokens are replaced with placeholders before the request leaves your machine and restored inside the incoming response. An optional 1.6 GB local CPU model handles names and unstructured personal details.
How to get started
Magpie runs on macOS, Windows, and Linux as a menu bar app, desktop window, TUI, or web interface, with additional support for Docker and Termux. Mac builds are signed, notarized, and auto-updating.
To install, download the app directly from usemagpie.ai, or run the quick script in your terminal: curl -fsSL https://usemagpie.ai/install.sh | sh
Once installed, add your API keys or subscriptions under Providers, pick models for your installed agents under Agents, and you are ready to go. The masking feature is turned off by default in privacy settings and can be toggled on when needed.
Why it matters
For developers juggling multiple coding agents, varied API quotas, and subscriptions, Magpie removes the friction of manual reverse proxies and constant config tweaking. It brings order to a fragmented developer ecosystem.
Keep in mind that aggressive masking will prevent models from inspecting actual token strings, which can get in the way when debugging credential errors unless you whitelist them. If you only ever use one agent on its official subscription, you probably do not need it.
Sources (1)
- github.com(opens in a new tab)Originalgithub.com