Polywise

Polywise is a self-hosted system for people who want AI agents to work with saved research, notes, and project context. It keeps that material available across conversations so you don't have to explain the same work each time.

• Content and retrieval: Capture links and other content, organize it by project or agent, and let agents find relevant material through keyword, vector, and knowledge graph search.
• Agent groups: Give agents their own prompts, tools, skills, and knowledge, then bring specialists into a shared thread to compare responses or coordinate work.
• Persistent memory: Background processing extracts knowledge from conversations, while retrieval feedback strengthens useful connections and graph pruning adjusts what the system retains.
• Recurring workflows: Combine sessions, todos, schedules, and link fetching to automate repeated research or project tasks.

Polywise is open source under the MIT license. It runs as a local service with a browser interface, and also has desktop and command-line interfaces. Content, memory, and runtime state stay in the system you control.

You can use local llama-based runtimes or connect external model providers. Where model processing happens depends on that choice; local operation doesn't make requests to an external provider stay on your machine. Saved-content retrieval also requires embedding and reranking models.