Your PandaNpc account identifies your clients and host machines, stores account-level configuration, and controls which devices may connect. It does not replace the Claude, ChatGPT/Codex, or API-provider account used by a coding engine. Those are separate services with separate usage rules.
The easiest way to avoid billing surprises is to identify the execution path before a session starts: remote coding agent, PandaNpc built-in chat, or bring-your-own-key chat.
Sign in and use the same account everywhere
PandaNpc supports:
- Email verification codes.
- Google sign-in.
- GitHub sign-in.
Use the same PandaNpc account on web, desktop, and iOS if you want those clients to see the same connected machines and Agent Center configuration. Signing in with a different identity can create a separate account even when it belongs to the same person.
When PandaPaw is installed on a host, pandapaw login opens a device-authorization flow. Approval creates device-scoped credentials for that machine; you should not copy a general account token into config.toml. See the Quick Start Guide for the supported pairing flow.
Who pays for each kind of AI usage
| What you use | Where it runs | Which allowance or bill applies |
|---|---|---|
| Remote Claude Code | Your connected host | The Claude login, subscription, or backend configured for that local Claude Code process |
| Remote Codex | Your connected host | The ChatGPT/Codex login or provider configured for the local Codex process |
| Remote PandaCode | Your connected host | The provider and API key selected for PandaCode |
| PandaNpc built-in chat with a provided model | PandaNpc chat service | The current PandaNpc allowance, credits, or plan shown in the app |
| PandaNpc chat with your own provider | Direct from the signed-in client when a usable key and Base URL are configured | The external provider bills your API account |
Remote control does not make a provider request free. It changes where you supervise the session, not which model account the process uses. Always check the engine and model shown on the connection before handing it to another person.
PandaNpc plans and quotas
PandaNpc plans control service limits such as how many host machines or sessions can be online and which account features are available. Current prices and quotas are displayed in the Membership page and at checkout; those live values take precedence over screenshots or copied documentation.
AI-provider charges are separate. Upgrading PandaNpc does not automatically upgrade Claude, ChatGPT, OpenAI API, Anthropic API, or another model service. Likewise, paying for a model subscription does not change PandaNpc device or session limits.
Built-in chat and credits
Built-in chat can use models provided through PandaNpc without requiring you to paste an API key. The model selector and account page show what is currently available and whether a request uses an included allowance or account credits.
Model availability changes more frequently than documentation. Use Supported AI Models to understand the model families and selection rules, then trust the live selector for the exact current list.
Bring your own API key
You can add an OpenAI-compatible or supported provider under Settings → Model Configuration. When a provider has both a usable key and Base URL, PandaNpc chat can call it directly from the active client. The external provider—not PandaNpc—meters those requests.
Keep several boundaries in mind:
- The provider must allow requests from the device and client you are using. Browser CORS policy can block an endpoint that works from a desktop client.
localhostrefers to the viewing device. It reaches a local Ollama/LM Studio server only when that server runs on the same device as the client.- A configured context window or maximum output does not increase the provider’s actual model limit. Values that exceed the upstream limit can produce rejected requests.
- Treat API keys as secrets. Do not paste them into screenshots, support messages, shared sessions, or documentation examples.
Follow Custom Providers and API Keys for field-by-field setup and a safe test procedure.
Device and connection security
Review Connected Devices periodically. Remove or revoke machines you no longer own, and prefer one-time device keys for normal installation. A saved connection can expose powerful coding tools, so account access should be protected as carefully as access to the development machine itself.
Connection sharing is also charged to the machine owner’s engine or provider. Read the warnings in Share a Claude Code or Codex connection before sending a live-control link.
Data and privacy
Different features process different data:
- Coding tools and project files execute on the selected host machine.
- Session messages and status travel through the relay so authorized clients can display and control the session.
- Prompts, referenced content, and attachments may be sent to the model provider chosen for that session.
- Shared links deliberately expose the content or connection described in their confirmation screen.
Read the Privacy Policy for the authoritative description of collection, providers, retention, consent, and your rights. If a short product description conflicts with the policy, the policy controls.
Account help
For sign-in, billing, data-access, or deletion requests, email support@pandanpc.com from the address associated with the account. Support may need to verify ownership before changing account data. Never include verification codes, API keys, device keys, or model-provider passwords in the message.