NanoNets has introduced Ami — an open agentic framework designed to run a personal AI assistant directly on a user's local device while maintaining complete data privacy.
What Happened
Developers from NanoNets have released Ami, a local agent harness that integrates with various tools via personal tokens. The system analyzes user actions to build a knowledge graph, allowing the agent to mimic the owner's communication style and logic when performing daily tasks. All data is stored locally in SQLite and Markdown formats in the ~/.ami/ directory.
Context
The project utilizes the Anthropic Messages API protocol, ensuring flexibility and compatibility with a wide range of models — from proprietary Claude to local solutions running via Ollama or LiteLLM. This allows the computational load and context storage to be moved from cloud servers to edge devices.
Why It Matters for the Industry
Ami demonstrates the growing trend toward decentralized and private agentic systems, stimulating the development of the edge AI ecosystem. The use of standardized protocols reduces developer dependency on cloud platforms and simplifies the creation of privacy-first personal agent prototypes.
Why It Matters for Users
Users gain the ability to run their own digital twin to manage to-do lists, prepare for meetings, and automate routines. Meanwhile, sensitive information is not transmitted to third-party servers, as all knowledge and action history remain on the local computer.
What Is Not Yet Known / Limitations
For full production-ready use, additional data regarding latency, computational resource requirements, and overall system reliability is required.
Sources
Author
Look at AI, Editorial Team