DEDUPEDWho it is forDeveloper whose big model gets burned on simple questions like "what does this do?"
The problemEvery turn goes to the same expensive model, even trivial ones, which wastes quota.
The outcomeSimple questions go to a fast model while test-writing and heavier work go to the strong model.
What you need first
- macOS, Windows, Linux, Docker or Termux
- At least one coding agent already installed (for example Claude Code, Codex or Gemini CLI)
- At least one provider: an API key, a local model server, or a signed-in subscription such as Claude, ChatGPT, Copilot, Gemini or Grok
- Bun is downloaded automatically the first time a plugin needs it
Steps
- Create a routing group containing a fast model and a strong model
- Turn on intent routing for the group using the setup in the intent docs at usemagpie.ai/docs/intent
- Point your agent at the group and ask one quick question, then one bigger task
- Open the Routing view to see which model handled each turn
Prompt for your agent
Set up yetone/magpie intent routing so <FAST_MODEL> handles simple questions and <STRONG_MODEL> handles test writing. Confirm in the Routing view that two different turns went to different models.
Caveat. Intent routing needs a small model of your choice to classify each turn, which adds a call per turn. Exact setup steps are in the linked docs, not the truncated README.
About magpie
A menu bar app to pair any coding agent with any model, like Codex on DeepSeek. Part of this week's top 10 repos: ranked #9 on Repo Radar. View yetone/magpie on GitHub.
More with magpie
- Run your coding agent on a cheaper or different model Point Codex, Claude Code or Gemini CLI at DeepSeek, Kimi or GLM without hand-editing each agent's config.
- Keep coding when one account hits its rate limit Group several accounts or models so the gateway quietly moves to the next one when a quota runs out.
- Point any base-URL tool at one local gateway Aim scripts and apps that take an OpenAI, Anthropic or Gemini base URL at magpie's local address.
- Use a plan you already pay for inside other agents Let OpenCode or Gemini CLI run on the ChatGPT, Claude or Copilot subscription you are already signed in to.
- Switch every agent's model setup with one named profile Save a Budget or Focus profile and flip all your agents between them in one move.
- Move your existing provider setup into one app Import providers from CC Switch, Claude Code, Codex or Alma instead of re-entering every key.
- Add a model login through an npm plugin Install an OpenCode or pi plugin so a sign-in flow becomes a usable provider for your agents.
Disclaimer
Deduped lists open-source repositories that belong to other people. We do not own, maintain, audit, or vouch for any of them, and a repo appearing here is not an endorsement or a safety check. The use cases, steps, and prompts are written by AI and may be incomplete, outdated, or just wrong. Where a use case is marked "Seen in the wild", we link to public posts where someone described something similar, but we have not verified those posts and the use case itself is still our AI's own write-up. Use all of this at your own risk. Only run code, commands, or agents on systems and accounts you are authorized to use, and read the code and every command before you run it, especially anything that touches credentials, files, or the network. Nothing here is professional, security, legal, or financial advice. All trademarks and project names belong to their respective owners, and their use here does not imply affiliation. Content is provided as is, without warranties of any kind. Found a mistake, a safety problem, or something you want removed? Email [email protected] and we will look at it promptly. Full disclaimer page