Built a personal assistant that can write its own tools and subagents and can also explore on its own
retained text
### TL;DR: I built a personal AI assistant called MAVIS that has the ability to write its own tools and subagents and can also self-reflect to improve on what it knows. I seek feedback on the design. Edit: One of the design objectives is safety and it is not granted full freedom to do whatever it wants. I have been building a personal autonomous AI assistant that has the ability to adapt to user needs by building out its own tools and subagents. I have tried to follow the design objectives to make it as deterministic, secure and flexible as possible. Along the way, I have also made an experimental self-reflection that helps the agent pick up useful topics and explore them to improve its knowledge base. I have written a small blog post series about the project design where I have discussed the major components of planning, synthesis, security, observability, caching and self-reflection. I would like to get feedback on it. The links to blog and repo are in comment.
document fetched 2026-10-22 00:19 · canonical https://reddit.com/r/AI_Agents/comments/1xbm37c/built_a_personal_assistant_that_can_write_its_own