Docs · Guide

Mister Morph on any model

Mister Morph (the morph command) keeps its settings in ~/.morph/config.yaml, or the file MISTER_MORPH_CONFIG names: the main model under llm.model, asked of the vendor llm.inference_provider names at llm.endpoint with llm.api_key. It has no list of providers to add one to, so magpie takes those llm keys over, and Mister Morph runs on any model in magpie: DeepSeek, GLM, Kimi, Qwen, Claude or a ChatGPT subscription.

Set up Mister Morph

1. Add a provider in magpieAdd an API key on the Providers page, or sign in to a subscription.
2. Pick a model for Mister Morph on the Agents pagePick one via magpie, and an effort if you want. magpie finds it by its ~/.morph folder, not the morph command, which other tools are named too; the folder is made by its first run, morph install or the desktop app's setup.
3. Start a new taskMister Morph's Console uses the model for new tasks; restart open morph chats to use it there.

magpie morph deepseek/deepseek-v4-pro

config.yaml

magpie writes these keys under llm: inference_provider openai_response_compatible, OpenAI's Responses API, Mister Morph's own default, which the gateway serves for every vendor; endpoint the gateway's /v1 (Mister Morph adds no second /v1 to a base ending in one); api_key magpie's key; and model the catalog id. In Mister Morph the model reads magpie/deepseek/deepseek-v4-pro. magpie also sets a User-Agent of mistermorph in llm.headers, as its requests otherwise carry the OpenAI SDK's, which tells them from no one's.

# ~/.morph/config.yaml (or $MISTER_MORPH_CONFIG)
llm:
  inference_provider: openai_response_compatible
  endpoint: http://127.0.0.1:3425/v1
  api_key: 
  model: deepseek/deepseek-v4-pro
  context_window_tokens: …
  headers:
    User-Agent: mistermorph
  reasoning_effort: high

llm.provider is left as it is: Mister Morph derives it from inference_provider. A Mister Morph in WSL is set up the same way, at ~/.morph/config.yaml, as MISTER_MORPH_CONFIG isn't read there.

The context window

Mister Morph knows a model's context window only from a list of its own, by the id after its last slash. So magpie writes the window it knows in llm.context_window_tokens, and removes the key when it knows none.

Effort

The effort goes in llm.reasoning_effort: none, minimal, low, medium, high, xhigh or max. Unset, Mister Morph sends none.

magpie morph effort high

Profiles and routes

llm.profiles and llm.routes are yours, and magpie leaves them alone. A route that sends the main loop to another profile still does.

Switching back

Before it first takes llm over, magpie saves what you had under those keys. Switching back puts them back, and removes the ones you didn't have; when the User-Agent was alone in llm.headers, headers goes too. Picking one of Mister Morph's own models restores them the same way, then sets llm.model to it.

magpie morph default

FAQ

Can Mister Morph run on DeepSeek, GLM or a ChatGPT subscription?

Yes. Pick a via-magpie model for Mister Morph on magpie's Agents page, and magpie sets the llm keys in config.yaml.

Which API does Mister Morph use through magpie?

OpenAI's Responses API (openai_response_compatible), Mister Morph's default, which magpie's gateway serves for every vendor.

Why does magpie write context_window_tokens?

Mister Morph looks a window up only in its own list, which doesn't have magpie's model ids, so magpie gives it the window it knows.

Are my llm.profiles and llm.routes changed?

No. They are yours; magpie changes only its keys under llm and puts them back when you switch back.

More agents on the same models: Grok App on any model, fx on any model, Aside on any model and mu on any model. Something fails? See Fixing common errors, or ask on Discord.