The Private Agentic AI Platform

Who's This For?

Privateer is for everyone who refuses to choose between capable AI and their privacy. Your conversations are encrypted on your device or in our cloud, and every AI request runs through zero-data-retention providers that never store or train on your prompts. Frontier models and real agents, without giving up your data.

Chat Interface

Ask anything and get clear answers. Attach images, documents, and files to give every reply full context. Your conversations are saved securely, so you can pick up exactly where you left off.

privateer.pro

New: Stop Typing. Just Talk.

Tap the mic and have an actual conversation. Privateer hears where your sentence ends — no push-to-talk, no wake word, no button to hold — and starts speaking the answer while the rest of it is still being written. Talk over it and it stops mid-word to listen. It even learns your rhythm: how long a pause means you're thinking versus done, retuned every turn, so it stops cutting you off.

Your Data, Your Rules

Keep data in the cloud or fully on-device — all encrypted. Your data belongs to you, and only you. And you don't have to take our word for it.

• Encrypted on your device or in our cloud
• Zero-retention inference via trusted execution providers
• Names, emails & personal details redacted before the AI sees them
• Share a device and each account's on-device data stays sealed off from the others

Read the privacy policy · Verify it in the transparency layer

Private by Proof, Not Promise

Turn on Sealed mode and your prompt is encrypted on your device straight to a hardware secure enclave. Our relay routes and meters it, but can't read your prompts or replies — inference is served by Sealed providers like Tinfoil and Phala. And you don't have to take our word for it: your device verifies the enclave's hardware attestation on every response.

• Prompts encrypted on-device, straight to the enclave
• A blind relay — we route and meter, but can't read your traffic
• Client-verified hardware attestation on every response

Sealed — hardware-attested blind relay

Confidential-compute providers

Tinfoil Phala NEAR AI Intel TDX NVIDIA Confidential Computing AMD SEV-SNP

Whiteboard

Branch any conversation into a living board — explore ideas side by side, add images and notes, and pinch and pan to see how your thinking connects.

Access Multiple Models

Choose from hundreds of models across every major provider — OpenAI, Anthropic, Google and more. Search, filter, and switch anytime to find the right fit.

privateer.pro

Frontier chat models

GPT Claude Gemini Grok Llama DeepSeek Mistral Qwen Kimi K3

Generate, Edit & Animate

Turn a prompt into a finished image or short video in seconds — then edit it with AI. Just describe the change, dial in aspect ratio, resolution, and style, and refine until it's exactly right. Describe a sound and get that too — a door slam, rain on a window, a distant siren — ready to drop under a cut.

Image models

FLUX Nano Banana DALL·E 3 Ideogram Recraft Stable Diffusion 3.5

Video models

Veo 3.1 Kling Runway Gen-3 Luma Ray 2 Seedance Wan

Guide a video with pictures

Reference images Start & end frames Continue a clip Alter a clip

Sound effects

Stable Audio 3

A video doesn't have to start from words alone. Give the model reference pictures — a look, a character, a palette — and it works from them; or pin the first and last frame and let it fill the middle. Made a clip you like? Continue it from its final frame, or alter it: same clip, new direction. Those are separate choices on purpose. A model handed both a pinned frame and reference pictures renders the frame and quietly discards the references — at full price — so Privateer asks which you meant instead of guessing.

Sound effects are the one generator we can't route to a zero-retention provider, so we don't pretend otherwise: they're switched off until you turn on non-ZDR media generation, and your prompt is sent with nothing attached to your account. Everything else here stays on the default.

Turn a Picture Into a 3D Model

Hand Privateer an image and get back a real mesh — one you can spin on the screen, keep in your library, and take into a game engine or a 3D app. Give it up to four views of the same object and the model stops inventing the sides it cannot see. Don't have four? Privateer can draw them: it generates the front from your description, then makes every other view an edit of that same picture, so you get one object from four angles instead of four different objects.

• Ten models from five labs, each with its own controls
• The exact price for your exact settings, quoted before you generate
• Turn it around in the app, then download the file — GLB, or a textured OBJ where the model makes one

No two of these models take the same options — one is dialled by output resolution, the next by a texture tier, the next by a quality preset — so the studio draws its controls from the model you picked and re-quotes the price the moment you change one. Being straight about the trade, exactly as with sound effects: mesh generation runs on a provider we can't route to zero retention, so it stays switched off until you turn on non-ZDR media generation.

Cut It Together in Studio

Generating a clip is half the job. Studio is the other half: lay your clips on a timeline, trim them, crossfade between them, put a generated track underneath, and export the result back to your library. Drag in your own footage and it sits on the same timeline as everything the app made.

The editing itself never leaves your device: the media is decrypted in your browser, cut there, and re-encrypted before it is stored. There is no server-side render — nothing is uploaded to be processed. Generating inside a cut works exactly like generating anywhere else in Privateer, with the same disclosures. Studio needs a real browser engine, so it runs on the web app and the desktop app, not on phones.

Every Model, One Place

One account, every major provider. Chat with hundreds of models — GPT, Claude, Gemini, Grok, Llama, Mistral, DeepSeek, Qwen and more — plus image and video generation. Switch anytime in the app, or call any of them by ID through the OpenAI-compatible API.

Chat & reasoning · 40+ providers

OpenAI Anthropic Google xAI Meta Mistral DeepSeek Qwen Moonshot AI Z.ai MiniMax

Image & video

FLUX Nano Banana DALL·E 3 Ideogram Recraft Stable Diffusion Veo 3.1 Kling Runway Luma Seedance Wan

Browse all 300+ models → See API pricing →

Now Available: Kimi K3, Claude Opus 5 & GPT-5.6

The newest frontier models are live on Privateer: Moonshot AI's Kimi K3, Anthropic's Claude Opus 5 and Claude Sonnet 5, and OpenAI's GPT-5.6 line — Luna, Sol and Terra, plus their Pro tiers. All of them are built for sharp agentic reasoning and standout coding, and you can switch between them mid-conversation. Ask one to design a game and save the result as Cargo: a playable build stowed encrypted in your hold, ready to play, tweak, or hand to a friend through a private link. Chat with them as privately as every other model — encrypted on your device or in our cloud, with zero-data-retention inference.

Newest from Moonshot AI, Anthropic and OpenAI

Kimi K3 Kimi K2.7 Code Claude Opus 5 Claude Sonnet 5 GPT-5.6 Luna GPT-5.6 Sol GPT-5.6 Terra GPT-5.5

Ask any of them for a game — save it as Cargo

Arcade shooter

Puzzle platformer

Card game

Browse all 300+ models →

Build & Save Cargo

Ask for something real — a web page, a slide deck, a game, a document, or a spreadsheet — and save the result as Cargo. Every build is stowed encrypted in your hold, ready to run, edit, or download anytime — or share it with anyone through a private link. A file you already have can become one too: a spreadsheet or document you attach, or keep in a project, carries an Edit badge that turns it into an artifact you can keep updating, with version history. The conversion runs on your device, so it tells you what a format change would cost before anything is written.

Code in Your Terminal

Privateer comes to your command line. The CLI is a coding agent that reads, writes, and runs code across your projects — signed into your account, so there's no API key to manage. Link a terminal to the app to drive it from your phone, approve its actions, and send it files.

Command Your Fleet of Agents

Every terminal you link becomes an Agent you can drive from your phone or the web. Send it a prompt, watch it work in real time, and approve each action — or give no quarter and let it run to the finish. Bridge it into your team chat, too — Telegram, Slack, Discord, or WhatsApp — and message it like a teammate, approving its actions right in the thread. It also speaks the Agent Client Protocol, so an ACP host like Buzz or Zed can drive the same agent — still bound by the tool limits you set and confined to the directory you point it at. It sends finished files straight back to you, and you can take over driving it from any device.

Work that finishes while you're away lands in your inbox: what the run produced, the files it staged, and a button to have it read to you — long results included, read in full rather than cut off partway. From a result you can follow up, which starts the next task on the machine that produced it and carries that routine's own instructions and working directory with it; if that machine is offline, Privateer asks rather than quietly running it somewhere else. Or ask for a brief: one short read across everything you haven't seen yet, with a line on why each item matters. Briefs are written on your device and stay there.

Then Run It All By Voice

Voice mode isn't only for chat. Point it at a linked terminal and drive a real coding agent hands-free: hand it the job out loud, hear back its one-line account of what it did, and approve tool calls by saying so. Say allow, or deny. Say "no, don't allow that" — both words in one breath — and deny wins, because a wrongly-denied command costs you a re-ask and a wrongly-allowed one already ran on your machine. Anything it isn't sure it heard, it asks again. It never guesses.

Privateer on your desktop

A native desktop app for macOS and Windows — the full Privateer environment in its own window, with a local agent that can read, write, and run code in a folder you grant it. Same account, same encryption; your keys never leave your machine.

Harbor — An Agent That Never Sleeps

Don't want to keep a terminal running? Harbor hosts your Privateer agent in our confidential cloud — always on and reachable from your phone or the web, with your MCPs, tools, credentials, and scheduled routines. It runs inside a hardware TEE we can't read into, the same posture as our Sealed inference. Included on Navigator and above.

Build on the Privateer API

An OpenAI-compatible developer API, billed to your Privateer account. Point your existing OpenAI client at our base URL, drop in a sk-priv-… key, and reach hundreds of chat, vision, image, video and audio models through one endpoint. Inference is a stateless pass-through — we persist only billing metadata, never your prompts or the responses.

Mobile & Web, In Sync

Privateer goes wherever you do. Use it on your phone or right in the browser — and with cloud storage, your encrypted conversations, projects, and boards stay in sync across every device.

privateer.pro

Questions, Answered

Straight answers to what privacy-minded people actually ask.