Skip to main content

Cumbersome: Free LLM API Client for iPhone and Mac

Bring your own API keys. Choose the exact model. Connect tools, compare answers, and control how every conversation works across iPhone, iPad, and Mac.

Face/Off Mode
Choose the Exact Model
Your API Keys

"A feature I have never seen before is the Face/Off mode... The developer built this app out of his own need, which in my experience is a good foundation."

One LLM Client
Your Models, Tools, and Data

Cumbersome is a native LLM API client for people who want more control than consumer AI apps provide. It connects your Apple devices directly to cloud APIs, local models, and tools without putting a Folding Sky server in the middle.

  • Use your own API access. Add direct provider keys, a multi-provider gateway, or your own OpenAI-compatible endpoint. You pay the provider for what you use instead of paying Cumbersome a subscription. Read why I prefer API keys over consumer AI subscriptions.
  • Choose the exact model. Cumbersome does not choose a cheaper or faster model behind your back. Switch providers or models in the same conversation and see which one produced every response.
  • Work on every Apple device. Conversations sync through your iCloud account. Start on an iPhone, continue on a Mac, and keep the same editable history without creating a Folding Sky account. See how to use LLM API keys on iPhone without relying on a browser playground.
  • Run local LLMs when you want them. Connect an OpenAI-compatible local endpoint and use models running on your own Mac from Cumbersome on Mac or iPhone. The Ollama setup guide walks through the complete setup.
  • Keep the controls. Edit any message, set system instructions, pass custom API parameters, manage reasoning, inspect token usage, and export the underlying data. Cumbersome exposes the controls consumer apps tend to hide.
  • Control long-running costs. Prompt caching exposes cache usage, while conversation compaction replaces old API context with a shorter summary without deleting the transcript. You can see what the API receives and decide when to reduce it.
  • Compare instead of guessing. Face/Off Mode sends one prompt to three models and has another model judge the responses. You still see every answer and make the final call.

More Than an LLM Chat Window

Bring Your Own LLM API Access

Connect direct provider keys, gateways, local LLMs, or another OpenAI-compatible endpoint. Choose the model yourself and switch it without starting over.

Local Browser and Remote MCP Tools

Let models search and browse through a local headless browser, or connect remote HTTPS MCP servers and choose which tools each conversation can use. See how local browser tools work.

Compare LLMs with Face/Off Mode

Send one prompt to three models in parallel, compare every response, and let another model judge the result. See Face/Off Mode.

Generate Images and Run Code

Generate images through supported models. Run HTML and JavaScript responses in an in-app preview instead of copying code into another tool.

Prompt Caching and Compaction

Inspect cache-token usage and compact long conversations into shorter API context. The original transcript stays visible while future requests send less history.

Folders, Presets, and Editable History

Organize conversations into synced folders, apply reusable system-message presets, and edit any message when the context needs correction.

Also Included

Totally free. No upsell, no subscriptions.
System-message presets and custom API parameters.
iCloud sync, Keychain storage, no Folding Sky backend.
Editable Markdown, LaTeX, attachments, and JSON export.
Siri and Shortcuts for hands-free use.
VoiceOver, Dynamic Type, full keyboard nav.

Built for People Who Want the Controls

Frustrated by

  • Paying several AI subscriptions to reach different models
  • Auto-routing that makes the model choice unclear
  • Browser playgrounds that are awkward on iPhone and iPad
  • Long conversations that repeatedly send expensive context
  • AI interfaces that hide tools, parameters, and raw usage

Built for

  • Developers who want an LLM API client without building one
  • Researchers who compare models on the same prompts
  • Professionals who move between iPhone, iPad, and Mac
  • People connecting local models, browser tools, or MCP servers
  • Power users who want visible costs and full parameter control
Bring your own LLM API access
Choose the exact model for each conversation
Full parameter control
Face/Off Mode compares three responses
Read practical guides for LLM APIs and Cumbersome →
Peter with dahlias

Built by Peter Bray. Bootstrapper in Reno. He sometimes takes on outside work.