VantaWhat you can do with Vanta
Use self-hosted, on-device and cloud AI models on iPhone and iPad, with documents, memory, group conversations and compatible tools.
Vanta brings your models, conversations and tools into one iPhone and iPad app. Connect a server you already run, download a supported model, or add a cloud API. You choose where each conversation is processed.
Take your server with you
Use models running in LM Studio, Ollama, vLLM or another OpenAI-compatible server. Your computer does the inference; Vanta gives you the mobile chat interface. A model does not need to fit in your phone’s memory when it runs on your server.
Connect LM Studio · Connect Ollama · Connect vLLM
Run a model on your device
Browse supported GGUF and MLX models from Hugging Face, or import a GGUF file. Downloaded models can run without a server or API key. Apple Intelligence is also available on eligible devices when the system model is ready.
GGUF is text-only in Vanta. Selected MLX models support images and tool calls; those capabilities depend on the model. Available memory, model size and context length determine what your device can handle.
Bring your own cloud API
Use OpenRouter or add an OpenAI-compatible provider endpoint with your own key. Save model presets, adjust supported generation settings and switch models from the chat picker. Provider usage is billed separately by the provider.
Give your chat something to work with
Attach PDFs and ask focused questions about their contents. Vanta extracts text and retrieves relevant passages when a document is too large to include in full. Use a compatible vision model to discuss photos and screenshots.
Add remote MCP servers to give compatible models access to documentation, search or your own tools. Built-in web search uses a separately configured Tavily key. Tools may contact external services even when your chat model runs locally.
Work with files · Connect MCP tools
Prompts, memory and context
Memory can carry relevant details between conversations. Review memories, adjust retrieval and automatic extraction, or turn extraction off. Long chats can be summarized to manage their context window, and pinned messages survive compaction.
Use saved system prompts and model presets for different tasks.
Multiple models, one conversation
Group conversations let multiple models contribute with assigned roles. Combine local, self-hosted and cloud participants, compare their answers and let them challenge each other’s ideas. The orchestrator uses a remote model.
Speak when typing gets in the way
Choose on-device Apple Speech or download a WhisperKit model for voice input. Cloud transcription is optional and uses the provider you configure. The choice of speech engine is separate from the choice of chat model.
A native reading and writing space
Read formatted Markdown and syntax-highlighted code in a native iPhone and iPad interface. Choose a theme and adjust the appearance to suit how you like to read.
Pick the workflow that fits
Start with one model. Add documents, memory and tools when they help the task. Compatibility explains the limits, and privacy shows where information goes in each mode.
Your next chat starts here.
One purchase. Your choice of models.