AI 에이전트 앱을 발견 할 수있는 능력을 제공, 제어를 이해, 작업 데스크탑에. 컴퓨터 사용 MCP는 MCP 호환 에이전트를 기본 도구에 연결합니다. ** 마이크, 윈도우 및 리눅스 **.

응용 프로그램 스크립트 및 액세스 제어를 사용할 때 사용할 수, 및 스크린 샷, 인터페이스가 필요할 때 마우스와 키보드 입력.

페인트 보기

Astra는 실제 마우스 스트로크를 통해 이것을 그린. OpenAI를 사용한 대리인 이 MCP를 통해 번들 페인트 스튜디오를 운영하는 응답 API: 105의 치기, 7 가지 색상과 15 가지 모델 호출, 검증 된 PNG 수출. 결과가 원래 Van Gogh-inspired 풍경은 네이티브 포인터 이벤트를 통해 만들었습니다.

이것은 유료 API 호출합니다. 기록 된 실행은 191,734 입력 및 3,274 출력을 사용 토큰. 데모를 실행하고 한계를 읽으십시오.

읽기

**Thirty-eight 초, 10 필드, 대답 키. ** 왼쪽의 송장은 캔버스에 그려, 그래서 그것의 텍스트는 접근성 나무에 결코 도달하지 — 유일한 방법 그것을 읽는 것은 봐. DeepSeek Flash는 스크린 샷, 각 값을 읽습니다. Accessibility API를 통해 Verify를 누르십시오. 로컬 호스트는 항목은 하나의 잘못 만 알려줍니다.

Computer Use MCP

Give an AI agent the ability to discover apps, understand their controls, and work on a desktop. Computer Use MCP connects MCP-compatible agents to native tools on macOS, Windows and Linux.

Use app scripting and accessibility controls when available, and screenshots, mouse and keyboard input when the interface requires them.

See it paint

Astra painted this through real mouse strokes. The agent used the OpenAI Responses API to operate the bundled paint studio through this MCP: 105 strokes, seven colors and 15 model calls, followed by a verified PNG export. The result is an original Van Gogh-inspired landscape made through native pointer events.

# From a clone of this repository, after the setup below:
npm install --prefix agents/openai-agent
npx playwright install chromium
export OPENAI_API_KEY=your-key
node agents/openai-agent/showcase.mjs paint --studio --turns 24 --tokens 200000

This makes paid API calls. The recorded run used 191,734 input and 3,274 output tokens. Run the demos and read their limits.

See it read

Thirty-eight seconds, ten fields, no answer key. The invoice on the left is drawn on a canvas, so its text never reaches the accessibility tree — the only way to read it is to look. DeepSeek Flash reads it from screenshots, types each value through the accessibility API, and presses Verify; the local host scores the entries and tells it only which ones are wrong. This run scored 10 of 10 on the first Verify, over 12 model calls and 23 tool calls, at a 94% prompt-cache hit rate. Every entry is recorded with its isTrusted flag, so a submission that was not typed through the UI is refused.

npm install --prefix agents/deepseek-agent
npx playwright install chromium
export DEEPSEEK_API_KEY=your-key
node agents/deepseek-agent/showcase.mjs --seed 7             # the run above
node agents/deepseek-agent/showcase.mjs --receipt random     # a real receipt from the web

Full-resolution capture · how it works, and what it cannot fake

See it plan

Watch the work instead of reading tool calls. The run console is an MCP App the agent reports into. It declares a plan, marks one task active at a time, and attaches real screen captures as it goes — the frame above is a live capture of Blender’s own window.

Underneath the conversation is everything it is doing: each thought, each tool call and each result, with timings. That feed is reported by the host driving the run