Install and sign in
Download PandaOS, answer the short onboarding and choose where its models come from.
How-to guide Updated 20 Nov 2026What you get
PandaOS installed on your computer, set up with your name, the level of detail you want in answers, and the AI provider Panda uses.
Install
- Download it from pandaos.ai/dashboard.
Pick your platform. - Run the installer and open PandaOS.
- Sign in.
Pick Continue with Google, Continue with Microsoft or Continue with GitHub, or get a sign-in link by email. - Type your name.
Panda uses it to greet you. A short thank-you screen follows. Press Continue. - Answer How technical are you?
It sets how much Panda explains. Change it later under Settings, Account. - Pick a provider under Choose your AI provider, then set it up.
If your account has PandaOS Cloud usage to spend, onboarding skips the list and opens Set up PandaOS Cloud. Press or use another provider to pick something else. The table under Which provider to pick says what each choice means. The last step checks that the connection works.
Which provider to pick
The choices sit under four headings: PandaOS, Subscriptions, API keys and Custom.
| Choice | What Panda uses |
|---|---|
| PandaOS Cloud | Models from several providers, run by PandaOS, with no keys to manage. Usage counts toward your plan. |
| Free Mode | The free tiers of several providers, such as Google, Groq and Mistral. |
| Claude | Your Claude subscription, through sign-in. |
| OpenAI Codex | Your ChatGPT sign-in or an OpenAI API key. |
| OpenCode Go | Your OpenCode Go subscription. |
| Anthropic with Agent SDK | Your own Anthropic key, paid per use. |
| Claude on AWS Bedrock | Claude through your own AWS account, with a bearer token and a region. |
| Your API Keys | Keys you paste from the providers you choose. You pay those providers directly. |
| Local & custom endpoints | Models running on your own computer, in Ollama, LM Studio, KoboldCpp or llama.cpp, or any OpenAI-compatible address. |
If you are unsure, start with PandaOS Cloud: it is first in the list and needs no keys. In an organization, your organization's cloud takes its place. Connections explains switching later.
What you see when it worked
A Welcome message with your name appears. PandaOS then creates a demo project and opens it. Your next step is Your first project.
If something goes wrong
My local model does not show up
PandaOS looks for each local app at its usual address on your computer. Start the app first, and keep its default address: Ollama at localhost:11434, LM Studio at localhost:1234, KoboldCpp at localhost:5001, llama.cpp at localhost:8080. During onboarding, pick Local & custom endpoints, then press Scan again once the app is running.
I picked the wrong provider
The provider is a setting, not a commitment. Open Settings and pick Models to choose another one. A single chat can also move to another connection whenever you like.
The answers are too technical, or not technical enough
Open Settings, pick Account and change Expertise. It sets how technical Panda's answers are.