Personas combine prompts, tools, TTS and memory slots into one persona as a preset.
Swap prompt parts so you can change parts of the prompt and build them like blocks.
Online store for plugins. In-app browser for the store.
Tell your AI which tools it is able to use and which are invisible.
Prompt injections that change each turn for more variety.
Say “Hey Sapphire” then start talking to the app naturally.
Kokoro TTS streams voice while LLM is generating. Swappable.
You speak to the app and it transcribes your voice to text.
The AI can edit its own prompt, name, tts voice.
The AI can make tools to interact with the world around you.
Give your AI a task to do every X minutes on repeat.
Have tasks run at a certain time each day. AI or human can schedule.
Your AI acts when something changes to save tokens, no need to heartbeat everything.
Trigger the app or your personas from other apps.
The AI can read images from the app if it is a vision capable model.
The AI can browse the web for images and display them in chat.
We support Comfy, stablediffusion.cpp, Flux.
Memory stores single facts that the AI wants to remember.
Your AI has a contacts list with details and ‘allow email’ option.
A task list the AI can create and cross off items as it goes.
Storage for bigger data. Users could put a whole book here, it gets chunked and searched by chunk.
The AI can do a web search and get 3 websites from a single query.
Search the internet via DuckDuckGo.
Tool based email, and daemon based replies for instant checks.
Access to gcal (Google Calendar) via API.
Smarthome control. Lights, scenes, cameras, thermostats, sensors.
Tools for Github that cover basic operations like PR and comments.
Your AI can have it’s own bitcoin wallet, send and receive.
Remote access to servers and other computers.
Bring your own avatar. See your persona move as it acts.
Agents carry out tasks in the background, issued by your persona.
The app can call the Claude Code command and the persona can request code.
Online store for new personas. Also found in-app.
Full help guide for users. Searchable, in-app help.
In-app video guide. ~5 hours of the dev livestreaming casually.
The AI can search its own help docs to understand its own systems.
We have a tiny RAG model for searches. Swappable via plugin.
Discord bot support. The AI can do it via tools or reply from daemon.
VAD knows when a user stops talking so it cuts off recording.
Clock, stopwatch, timer and alarm. A gentle ping noise.
Really simple scratch pad for your persona.