otaku
Free LLM roleplay client with a web UI and terminal; runs locally against a local model or cloud API.
Last verified:
What is otaku?
otaku is a free, open-source LLM frontend for interactive roleplay with web and terminal interfaces. It runs entirely locally without requiring Docker or database servers, and integrates with local models or cloud APIs with minimal configuration.
otaku pricing
Pricing model: Freemium
Free (open source, MIT license)
otaku pros
- Maintains consistent prose by always keeping first 20 and last 150 messages intact to preserve LLM style
- Automatic character extraction and tracking—each character maintains individual memories of story events
- Flexible roleplay without fixed personas—users can play multiple characters and hint identity to the LLM
- Import compatibility with SillyTavern chats, character cards, text files, and world info
- Local-first design with optional AES-256-GCM encryption at rest and API keys sealed in OS keychain
- Transparent context inspection via /context command to see exactly what will be sent to the LLM
otaku cons
- Context management via scene summaries may lose fine-grained details as conversations grow longer
- No built-in community character library or sharing features—limited discovery of community content
- Requires manual local setup and configuration compared to cloud-only alternatives
Frequently asked questions about otaku
Is my data private and secure?
Stories live in a local SQLite database with daily snapshots. Encryption at rest (AES-256-GCM) is optional but available with keys stored in your OS keychain or derived from a passphrase.
What LLM providers can I use?
otaku works with any local model or cloud API. Provider API keys are always stored encrypted, with the key in your OS keychain.
Can I import my existing roleplay content?
Yes—you can import SillyTavern chats, character cards, text files, and lorebooks/world info into your stories.
How does it handle long conversations?
As messages grow, intermediate messages are automatically split into scenes and replaced with summaries, while the first 20 and last 150 messages are always sent unchanged to preserve style.