The open source agentic content system
English | 简体中文 | 繁體中文 | 한국어 | Deutsch | Español | Français | Italiano | Dansk | 日本語 | Polski | Русский | Bosanski | العربية | Norsk | Português (Brasil) | ไทย | Türkçe | Українська | বাংলা | Ελληνικά | Tiếng Việt
Polywise is the open-source agentic content system. You can use it from the command line or the desktop app to chat with models, save knowledge, retrieve context, and turn repeated working styles into reusable agents.
Polywise has two practical entry points: the CLI and the desktop app.
Install the CLI globally:
npm install -g polywiseStart the local Polywise service:
polywise start
polywise start -dpolywise start keeps the service in the foreground. polywise start -d exits immediately and leaves the service running in the background.
Then visit web UI at http://localhost:3072/app/ .
You can enable Auth login within the settings. Once enabled and a password has been set, you will be required to log in when accessing the Web UI in order to access the API. This is crucial if you are deploying Polywise as a service on a server for remote access.
Download the latest desktop build from GitHub Releases.
The desktop app is the easiest way to explore sessions, saved content, agents, and posts without working from the terminal.
For a first run, Polywise mainly needs:
- one available model provider
- embedding and rerank models if you want saved-content retrieval
You do not need to configure every provider or integration on day one.
polywise upgradeInstall the latest release from GitHub Releases.
If you want the shortest path to first value:
- Open
Settings -> Model Providerand configure one provider you can actually use. - Open
Settings -> Model Settingand make sure the default chat model is available. - Go to
Sessionand ask one real question instead of sendinghello. - Save one short note, page summary, or answer into Polywise.
- Mention that saved item again in chat to verify retrieval.
Once one provider is connected and the default model is set, stop configuring and go use the product.
The app becomes easiest to understand when you use each area for one concrete job:
Sessionfor asking real questions, planning work, and staying inside your workspace contextLinkcasefor fetching and extracting web content into the systemAgentfor turning repeated instruction styles into reusable collaboratorsPostsfor saving knowledge that should live longer than a chat reply
Two shortcuts are worth learning early:
@brings files, agents, and other context into a session/brings tools and skills into the workflow
The CLI is a thin wrapper over the backend API. By default it talks to http://localhost:3072; set POLYWISE_SERVER_URL if your server lives elsewhere.
Start with help instead of memorizing commands:
polywise -h
polywise session -h
polywise session create -hUse input_schema when you need the exact input shape for a command:
polywise input_schema session.createCommon commands:
polywise start
polywise start -d
polywise version
polywise session create --title "Daily Review"
polywise search fullTextSearch --query "vector database"
polywise save --for user --content "Key takeaway..."When payloads become more complex, pass JSON directly:
polywise search fullTextSearch --json '{"query":"agent memory","for_types":["wiki","memory"],"enable_recall":true}'Polywise is built upon the belief that truly intelligent AI requires truly intelligent memory. It is not merely about storage, but rather a system capable of organically forming connections, strengthening through use, strategically forgetting, and continuously evolving.
This project was inspired by the following research papers:
- Long-lasting potentiation of synaptic transmission (1973)
- The Organization of Behavior (1949)
- A Spreading-Activation Theory of Semantic Processing (1975)
Polywise stands on the shoulders of these excellent open-source projects:
- 🐘 Sqlite - The most widely deployed high-performance embedded database in the world
- 🏹 sqlite-vec - Adds vector search support to Sqlite
- ⚛️ React - Front-end UI library
- 🖥️ Electron - Desktop application framework
- 🔗 tRPC - End-to-end typesafe APIs
- 📦 MobX - Simple, scalable state management
- 🎨 Tailwind CSS - Utility-first CSS framework
- 🚀 Hono - Ultra-fast Web framework
- 🛠️ Rsbuild - Next-generation build tool powered by Rspack
- 📚 Rslib - Library build tool powered by Rsbuild
- 🤖 AI SDK - Unified toolkit for building AI-powered applications
- 🤗 node-llama-cpp – Node.js bindings for llama-cpp, designed for interfacing with local models.
MIT – See LICENSE for details.