Utsuwa
utsuwa.ai
- Free plan
- Windows
- macOS
- Linux
- Web
- Android
- Self-hosted
What Utsuwa is
Utsuwa is a free, open-source AI companion app built around a 3D VRM avatar, voice, conversation memory and a relationship system. It supports models from OpenAI, Anthropic, Google, DeepSeek and xAI, along with OpenAI-compatible endpoints, Ollama and LM Studio. Users can load their own VRM avatars. Local embeddings enable semantic recall, and users can browse, search, add and delete memories. Voice options include hosted and local services; listed capabilities include voice conversations, proactive messages and image or video chat. Utsuwa can run entirely on a user's machine with local models and voice services. Its maker says there are no accounts, analytics, ads or tracking cookies, and that conversations, memories and settings are stored on the device. In the hosted web app, cloud-provider requests may pass through Utsuwa servers; the maker says those servers do not store messages or API keys or log conversation content. In the desktop app, messages go directly to the selected provider. The desktop app is in beta, and its builds are not code-signed. The maker says Utsuwa is intended for adults aged 18 and over.
Who it suits
It suits adults who want an AI companion with customizable avatars, voice and persistent memory. People who prefer local processing can use local models and voice services on their machine.
Good to know, and what to watch
- Supports a range of hosted and local model options.
- Users can manage stored memories.
- Can run with local models and voice services.
- Maker says it has no accounts or tracking cookies.
- Desktop app is in beta.
- Desktop builds are not code-signed.
- Intended for adults aged 18 and over.
Plans
Every plan and priceFree
Free plan
The verdict
Utsuwa combines an avatar-based companion with voice and memory, and can operate entirely on a user’s machine. Consider the desktop beta status and data routing differences between web and desktop before choosing how to use it.
What the maker publishes
- Age
- The maker states that Utsuwa is intended for adults aged 18 and over. source
- Avatars
- Users can load their own VRM avatar models. source
- Desktop data routing
- In the desktop app, messages go directly from the device to the selected provider without passing through Utsuwa's servers. source
- Desktop status
- The desktop app is in beta, and its builds are not code-signed yet. source
- Integrations
- The docs describe MCP tool connections including Home Assistant, Brave Search, and GitHub. source
- Memory
- Local embeddings support semantic recall of conversations, and users can browse, search, add, and delete memories. source
- Models
- It supports OpenAI, Anthropic, Google, DeepSeek, xAI, OpenAI-compatible endpoints, Ollama, and LM Studio. source
- Offline use
- Utsuwa can run entirely on a user's machine with local models and local voice services. source
- Platforms
- The browser app runs on desktop and mobile, while desktop builds are available for macOS, Windows, and Linux. source
- Privacy
- The maker says there are no accounts, analytics, ads, or tracking cookies, and conversations, memories, and settings are stored on the user's device. source
- Product
- Utsuwa is an open-source AI companion app with a 3D VRM avatar, voice, memory, and a relationship system. source
- Support
- The maker directs users to GitHub as the fastest way to reach the team. source
- Voice
- Voice options include ElevenLabs, OpenAI TTS, local TTS, OmniVoice, local Whisper, Groq, OpenAI, and browser speech recognition. source
- Web data routing
- In the hosted web app, cloud provider requests may be relayed through Utsuwa's servers, which the maker says do not store messages or API keys or log conversation content. source
Checked against the maker's pages on 5 Oct 2026.