Start AI in your browser without an installer or app server
Using AI should not require giving an application more access to your PC just to start it. Naidan’s Standalone edition runs from local files in the browser, keeping the browser’s boundaries without a separate application server.
Keep the application as a folder you can open. Model preparation, saved conversations and any local inference engine are separate parts of the setup.
Start from the downloaded folder
Download and extract
Download naidan-standalone.zip and extract the whole archive. Keep the directory structure, including assets, intact.
Open index.html
Open index.html in your usual browser. Its address starts with file://. Keep the supporting files next to it; the entry HTML is part of the extracted application, not a replacement for the entire archive.
Choose a model or engine
Use a supported in-browser model or connect an engine such as Ollama or llama.cpp server. Fetching model files may need a connection. A local inference server runs separately with its own process permissions.
After opening Naidan, follow the in-browser model guide for a GGUF or another browser model, or the local engine guide to connect a separate inference server.
What “no server” means here
No server to serve the application
Opening the files directly replaces the need to run an HTTP server for the Naidan interface. It also avoids choosing a listening port for that server.
The inference engine is your choice
An in-browser model runs inside the browser. Connecting a local engine is another option, with its own setup, model support and performance. Running from file:// does not automatically choose either mode.
Browser capabilities still matter
File URLs have different origin rules from HTTPS pages. A connection endpoint may need to allow the file page’s origin, and some browser capabilities work differently. Follow Naidan’s availability notices. The hosted edition is an alternative when a feature needs an ordinary web origin.
Choose how to retain conversations
Browser storage and the extracted application folder are separate. The Ephemeral option keeps Naidan conversation data in application memory rather than writing it to persistent storage. Reloading then ends that session; downloaded files, model caches and OS memory management are separate.
Keep a backup before changing environments
Export your data before switching browsers, moving between hosted and file-based use, or replacing your installation. Download and extract a fresh Standalone folder for an update, keeping its own assets together.