The three surfaces
The full Privateer workspace on your own computer: chat, media generation, projects, and a coding agent you can point at any folder. The agent runs inside the app on your machine.
privateer.pro/download/mac
privateer.pro/download/windows
Desktop guide
A private AI coding agent for your terminal. Provider-agnostic — bring your own model or sign in and bill your Privateer account. Open source, MIT.
curl -fsSL https://privateer.pro/install.sh | sh
CLI guide
OpenAI-compatible inference billed to your account — chat, vision, images, video, and audio. Point any OpenAI SDK at it; no new client library.
https://api.privateer.pro/v1
API reference
There is also a mobile and web app and a browser extension. Those are covered in-app and on the marketing site rather than here.
Which one do I want?
| If you want to… | Use |
|---|---|
| Chat, generate media, and organise work in a window | Desktop app |
| Have an agent read, write, and run code in a project folder | Either — CLI in a terminal, or the desktop app's Spawn Privateer at… |
| Drive the agent from an editor like Zed, or from Buzz | CLI over the Agent Client Protocol |
| Run scheduled or unattended jobs that survive closing the terminal | Local Harbor (CLI or desktop app) |
| Call models from your own program or backend | Developer API |
| Give the agent tools — GitHub, Linear, Postgres, your own server | Connectors (MCP), in the CLI or the desktop app |
The desktop app and the CLI are not alternatives you have to choose between — they read the same ~/.privateer home on one machine. Sign in once, and your account, model choice and connectors are there in both. The desktop app can even install the privateer command for you.
Privacy, honestly
Privateer encrypts your content on your device or in our cloud, with inference on hardware-attested enclaves. The three surfaces have genuinely different postures, and it is worth knowing which one you're in:
| Surface | Where your content lives |
|---|---|
| Desktop app & CLI | Files and history stay on your own disk under ~/.privateer/. Running a model sends that turn's prompt to the provider you chose. |
| The app's synced chats | Encrypted on your device; the server stores ciphertext. Generating a reply processes the request transiently through the server and provider under zero-data-retention terms. |
| Developer API | Not encrypted, by nature. You send prompts to our server, which forwards them to the provider. What we guarantee is retention, not secrecy in transit — a stateless pass-through that persists only billing metadata, never your prompts or the responses. |
Where a model runs inside a Trusted Execution Environment, the app can verify the attestation itself and marks it Verified. Routing we assert but your client did not check never earns that badge — the shield legend spells out the evidence behind each rung.
Getting help
Each guide ends with a troubleshooting section for that surface. Beyond those:
- privateer.pro/support — contact form and the help chat.
- The CLI is open source (MIT): github.com/privateer-agent/privateer-agent. The app server is not open source; what we publish is a transparency layer you can inspect.
- In the desktop app, Help → Report a Bug… and Help → Open Logs Folder.