Inno Agent

A self-hosted AI learning agent with editable learner profiles, a knowledge wiki and scheduled review. Runs on macOS, Windows or a private server.

Screenshot of Inno Agent website

Inno Agent is a personal AI learning agent that keeps track of your goals, misunderstandings and study materials across conversations. It's for individual learners and educators who want ongoing tutoring with records they can inspect and correct. The open-source project uses the MIT license and runs as a macOS or Windows desktop app, a self-hosted web interface or a terminal app.

Its memory separates your learner profile, a knowledge wiki and conversation history. You can edit the profile. The wiki turns PDF, Office, image and text materials into pages with source references, concepts and analysis, so you can check where an explanation came from. Conversation recall lets the agent use earlier discussions without folding every temporary exchange into your permanent profile.

Scheduled tasks handle spaced review, daily reflection and knowledge-base updates, with reminders through Feishu, WeChat or QQ. For programming study, Practice Lab pairs generated exercises with a browser terminal and saved run output. The agent can use actual errors and results when explaining what to try next.

Data stays in local files by default. Model choice is separate: Ollama and local models are supported alongside OpenAI, Anthropic and DeepSeek through compatible endpoints. Choosing a cloud provider sends model requests to that service.

Skills and workspace presets cover guided tutoring, essay feedback, lesson planning and interactive geometry. VeryMath workspaces add mathematical computing, paper study and Lean 4 proof assistance. The system is designed for one learner per instance; it doesn't provide multi-user authentication or tenant isolation.

Similar to Inno Agent