Dive

Dive is a desktop application that runs entirely on your machine, offering a secure graphical interface for Ollama and Pi agentic AI. Its standout feature is an ultra-compact offline Retrieval-Augmented Generation (RAG) engine that can index massive book libraries and enable the agent to answer questions without an internet connection. It also integrates with the Model Context Protocol (MCP), allowing agents to interact with external tools and APIs, and includes a sandboxed environment for executing custom scripts.

Strengths: fully local operation keeps all data private; offline RAG ensures availability and saves bandwidth; MCP integration extends agent capabilities; open-source and free, with full code transparency.

Limitations and risks: requires users to manually set up Ollama and Pi agents, which can be challenging for non-technical users; indexing large libraries may demand significant local resources; sandboxed scripts can introduce security risks if not configured carefully; the project is driven by a single community contributor, so features and documentation may be immature.