Sotto
github.com
- Free plan
- Windows
- macOS
- Linux
- Web
- iPhone and iPad
- Android
- API
- Self-hosted
What Sotto is
Sotto is free, open-source, self-hosted language-learning software built around CEFR courses and private practice. Its graded lessons cover grammar, reading, listening, speaking, and writing, with progression tied to demonstrated mastery. Ungated practice remains available, while an SM-2 vocabulary and grammar memory graph schedules reviews and helps shape adaptive listening. Learners can create classes from articles, papers, YouTube links, or personal interests; Sotto levels content to a CEFR course and provides citations. Configurable AI and speech providers include OpenAI, Anthropic, Ollama, faster-whisper, and Kokoro, with Gemini Live for spoken translation. A documented setup can run language models, speech recognition, and speech generation offline on the learner’s hardware. The application, database, and files stay on the operator’s stack, while cloud-sent content follows provider terms. Access includes a browser PWA, desktop launcher, terminal client, and iPad workbook PDF. It is free to run under AGPL-3.0, but users pay configured providers directly.
Who it suits
Sotto suits learners wanting structured CEFR study and private rehearsal before speaking with others. It also fits operators who prefer self-hosting, offline components, and configurable AI or speech providers.
Good to know, and what to watch
- Covers grammar, reading, listening, speaking, and writing
- Adaptive reviews use an SM-2 memory graph
- Builds leveled classes from personal sources
- Supports offline language and speech components
- Offers PWA, desktop, terminal, and workbook access
- Users pay configured AI and speech providers
- Cloud-sent content follows provider terms
- Managed hosted offering is only planned
Plans
Every plan and priceSelf-hosted
Free plan
The verdict
Sotto provides a broad, self-hosted learning stack with structured courses, adaptive practice, and multiple clients. Its free license is useful for operators comfortable managing infrastructure and provider costs.
What the maker publishes
- Access controls
- Self-hosted setup uses a shared password and local learner profiles, with optional WebAuthn passkeys that an administrator can revoke. source
- Adaptive practice
- A per-course vocabulary and grammar memory graph uses SM-2 spaced repetition to schedule reviews and seed adaptive listening. source
- AI and speech providers
- Sotto supports provider-selected AI and speech, including OpenAI and Anthropic API keys, local Claude Code or Codex agents, and local Ollama, faster-whisper, and Kokoro components. source
- Clients
- Sotto offers a browser-installable web app/PWA, a terminal client, and a desktop launcher for macOS, Windows, and Linux; Windows setup uses WSL2. source
- Costs
- The software is described as free to run under AGPL-3.0, while users pay their configured AI and speech providers directly. source
- Course skills
- Its graded classes cover grammar, reading, listening, speaking, and writing, with progression gated by mastery. source
- Course structure
- Its course covers grammar, reading, listening, speaking, and writing, with progression gated by demonstrated mastery. source
- Credential security
- The maker says per-account API keys are encrypted at rest using BYOK_ENCRYPTION_KEY. source
- Credential security
- The README says provider API keys are encrypted at rest with BYOK_ENCRYPTION_KEY and describes a networkless sidecar for refreshing local CLI credentials. source
- Data and privacy
- Sotto runs its application, database, and files on the operator's stack; content sent to configured cloud providers is handled under those providers' terms. source
- Data control
- Users run the application, database, and files on their own stack; content sent to configured cloud providers is subject to those providers’ terms. source
- Desktop and mobile access
- The maker documents a Tauri desktop launcher for macOS, Windows, and Linux, plus an installable browser PWA for phones, tablets, and laptops. source
- Free plan
- Yes source
- Integrations
- The maker documents Gemini Live API for real-time spoken translation and an MCP server for connecting Claude Code or Codex workflows. source
- Intended audience
- The maker describes Sotto as private rehearsal for learners who want to practice before speaking with a teacher, tutor, classmate, or stranger. source
- iPad workbook
- The web app generates class workbook PDFs for annotation in GoodNotes or another app, and the README says the native iPad app targets iPadOS 18+. source
- License and cost
- The project is licensed under AGPL-3.0 and says it is free to run, modify, and redistribute under the same terms; users pay their AI and speech providers directly. source
- Mobile apps
- Yes source
- Mock exams
- Practice exams follow Goethe, DELE, Cambridge English, or generic CEFR formats and provide mock feedback, not official scores. source
- Offline lessons
- Yes source
- Offline use
- Sotto documents a fully offline setup in which the LLM, speech recognition, and speech generation run on the user's own hardware. source
- Offline use
- The project documents an optional fully offline setup using a local LLM, speech recognition, and speech generation on the user’s hardware. source
- Other learning surfaces
- Sotto provides an iPad workbook PDF for annotation and a Rust terminal client with audio playback and microphone recording. source
- Personalized content
- Learners can build classes from articles, papers, YouTube links, or their interests, with content leveled to their CEFR course and citations provided. source
Checked against the maker's pages on 13 Oct 2026.