Getting started
Overview
Synonode puts the model on your own computer to work: it builds apps, writes and fixes code, answers, researches and writes documents, and checks its own work. You reach it from any browser or your phone.
Synonode has two parts. Synonode Runner runs on your computer, next to your model: it keeps your projects, runs the tools and checks the work. synonode.com is the app you open in any browser, on a computer or a phone. It logs you in and passes messages to your runner.
What you can do
Build apps and write code
Ask for an app in your own words. Synonode asks what it needs to know, writes a plan, shows you two designs to choose from, then builds it in the preview while you watch. It also writes scripts, fixes the code you already have, and works in any language your computer can run. See Plans, designs and builds.
Ask questions and learn
A question gets an answer, not a build. Synonode explains step by step, draws a diagram or a small interactive widget right in the chat, and writes a demo when trying something beats reading about it.
Research and write documents
Research mode searches and reads many sources and writes a report with every claim linked. Synonode writes Word, Excel, PowerPoint and PDF files, and searches inside the documents you give it. See Research and documents.
Trust the result
After a change, Synonode opens the page in a real browser and uses it the way you would, at computer and phone width. At the end of a build it goes through every page, button and form, and holds the result against what you asked for. See How it checks its work.
Work from anywhere
Start something at your desk and follow it from your phone. Talk to it while it works, queue the next message, dictate, have replies read aloud, and get a notification when it's done. See Phone, voice and notifications.
Make it yours
It remembers how you like to work, across every project. Teach it skills, connect your services with connectors, and choose what it may do without asking in permissions.
What you need
- A computer running Linux (x64 or ARM) or Windows 10 or 11. On Windows it runs on Windows itself, as Claude Code does, and needs no administrator rights. macOS isn't supported yet.
- A model: llama.cpp, LM Studio, Ollama, vLLM or Jan on that computer, or any OpenAI-compatible API with your key. Coding models do best. See Models.
- git, because every project is a git repository.
- An account, approved by an admin.
New here? Start with the Quickstart. Press Ctrl K anywhere in these docs to search them.