Screenshots
A look at the app.
Point AgenticPocket at a model that runs on-device, or connect a remote server. Chat, generate images, speak, and draw on the knowledge you give it. Free, with no accounts and no telemetry.
On-device or remote, your choice. Unless you connect a remote server, nothing leaves the device.
Download a GGUF model and run inference on your phone with llama.cpp. Nothing leaves the device.
Connect any OpenAI-compatible server (Ollama, LiteLLM, LM Studio, LocalAI, or your own) and use its models.
The assistant can use a calculator, look up your knowledge base, and call tools you connect (MCP servers such as Linear, Notion, GitHub).
Generate images on-device or via a connected server, and share them.
Speak to the assistant (on-device speech-to-text), hear it back (text-to-speech), and ask about a photo you capture.
Add your own documents and the assistant draws on them. Create personas with their own character, voice, and memory.
A look at the app.
There are no accounts, telemetry, or ads. Everything is stored on your phone. Export your settings, profiles, knowledge, and configurations to a single file (optionally password-protected) and restore them on another device or after a reinstall.