Skip to content

Frequently asked questions

The app is open source under the Apache License 2.0. You pay your model provider, search service, or gateway for usage. There is no PP Agent subscription and no PP Techs account.

The interface, your documents, and local tools work without a connection. A model call needs a reachable provider or gateway, so most work needs the internet unless you run a local model behind an OpenAI- or Anthropic-compatible endpoint.

Anthropic, OpenAI, Google, OpenRouter, Mistral, Groq, and the other providers built into Pi, or your own OpenAI- or Anthropic-compatible gateway. See Models and providers.

No. There is no PP Agent service in the request path. Requests go from your computer to the provider, search service, or MCP server you configure. See Security and privacy.

That is the point. A tech team configures models, tools, MCP servers, and agents once. Everyone else works in a chat window with plain-language settings and an approval step.

Yes, with limits. Saving files, shell commands, and changes in business systems ask first by default. Shell commands are off by default. Every tool can be set to automatic, ask, allow, or off.

Yes. Installers are published for macOS (Apple silicon and Intel), Windows x64, and Linux AppImage and .deb. Computer use is macOS and Windows only; the HTML preview is a desktop feature.

Export an agent as a .json file. It carries the agent’s rules, tools, skills, model, starter prompts, and folder setting. Review imported agents before use, since they can reference tools and paths you did not configure.

PP Agent includes an iOS and Android app built around the web app, and remote access serves a small-screen layout from the computer running the daemon. Both connect to your own computer; there is no cloud sync.

The app checks on start and under Settings › General › Updates. Windows, Linux AppImage, and Developer ID-signed macOS builds update in place. See Install.

github.com/pp-techs/pp-agent. Contributions are welcome; read CONTRIBUTING.md first.