# AI use cases for Nuphus MCP (mrpulor-gh/nuphus-mcp) Repo: [mrpulor-gh/nuphus-mcp](https://github.com/mrpulor-gh/nuphus-mcp) | Use cases: 4 | Page: https://dedupedai.com/use-cases/mrpulor-gh-nuphus-mcp/ nuphus-mcp is a Rust stdio MCP server that lets AI agents see the screen, control windows, mouse and keyboard, and drive Chrome or Edge. This file holds every use case of this repo. Each one starts with a "## " heading and ends with its URL. - Automate a signed-in website with one saved browser profile: https://dedupedai.com/use-cases/mrpulor-gh-nuphus-mcp/log-in-once-automate-web-tasks/ - Drive an existing anti-detect browser window over CDP: https://dedupedai.com/use-cases/mrpulor-gh-nuphus-mcp/drive-anti-detect-browser-over-cdp/ - Operate a native desktop app by its accessibility tree: https://dedupedai.com/use-cases/mrpulor-gh-nuphus-mcp/operate-native-app-by-accessibility-tree/ - Combine a vision model with local OCR for exact clicks: https://dedupedai.com/use-cases/mrpulor-gh-nuphus-mcp/vision-ocr-precise-desktop-clicks/ ## How to automate a signed-in website with one saved browser profile with Nuphus MCP Repo: [mrpulor-gh/nuphus-mcp](https://github.com/mrpulor-gh/nuphus-mcp) | Difficulty: Intermediate | Time: 45 minutes | Basis: From the README Log in once in the managed Chrome profile, then let an agent run browser tasks that reuse that session. **Who it is for:** An operator or analyst who repeats the same task on a website they already use and wants an agent to handle it. **The problem:** Repeated browser work is slow by hand, and re-logging in for every agent run adds friction. **The outcome:** The agent navigates, reads, and extracts from the site using the saved session in the managed profile. ### Steps 1. Install the Rust stable toolchain and build Nuphus MCP from source with Cargo. 2. Add the Nuphus MCP server to your MCP client as a stdio server. 3. Let the server open its managed Chrome, go to the target site, and log in once. 4. Ask the agent to use the browser_* tools to navigate, take a snapshot, and extract the data you need. ### Prompt to paste into your agent ```text Use the nuphus-mcp server to open in its managed browser, take a browser snapshot, and confirm the account is still signed in. Then extract from and return it as a list. Verify the snapshot shows the expected account name before you extract anything. ``` ### Caveat Since Chrome 136, remote debugging is ignored on the default profile, so keeping your real extensions and logins requires a copied profile or the managed profile. Risk, credentials: The managed profile keeps your signed-in sessions on disk, so only log in to accounts you own and protect the profile directory. Dual-use notice: only use this on software, sites and systems you own or are authorized to test. Disclaimer: Third-party repos we don't own or vouch for. AI-generated use cases can be wrong: review everything and run at your own risk. Full text: https://dedupedai.com/disclaimer/ URL: https://dedupedai.com/use-cases/mrpulor-gh-nuphus-mcp/log-in-once-automate-web-tasks/ ## How to drive an existing anti-detect browser window over CDP with Nuphus MCP Repo: [mrpulor-gh/nuphus-mcp](https://github.com/mrpulor-gh/nuphus-mcp) | Difficulty: Advanced | Time: 60 minutes | Basis: From the README Attach the server to a fingerprint browser you already run, with optional port self-healing on reopen. **Who it is for:** A team member who already runs an anti-detect browser and wants an agent to operate inside one of its windows. **The problem:** Fingerprint browsers often change debug ports when a window reopens, which breaks a fixed connection URL. **The outcome:** The agent attaches to the external browser through CDP and runs browser_* tools in that window, without launching a managed Chrome. ### Steps 1. Start your fingerprint browser with a debugging port: `chrome --remote-debugging-port=9222 --user-data-dir=...` 2. In the MCP client env, set `NUPHUS_MCP_BROWSER_CDP_URL` to `http://127.0.0.1:9222`. 3. (optional) Set `NUPHUS_BROWSER_EXE_PATH` and `NUPHUS_BROWSER_USER_DATA_DIR` so the server can re-resolve the port after a window reopens. 4. Call a browser_* tool from the agent and confirm it attaches to the external window. ### Prompt to paste into your agent ```text Use the nuphus-mcp server to attach to the external browser at http://127.0.0.1:9222 and take a browser snapshot of the current page. Return the page title and the first five interactive elements. Verify the snapshot comes from the attached window and not a managed Chrome. ``` ### Caveat If the external endpoint stops answering and no browser identity is set, attach failures are hard errors, and the server never falls back to a managed Chrome or kills your browser. Risk, account pooling: Running many accounts through anti-detect browsers can break site terms of service, and the README does not address this, so check the rules of each site first. Dual-use notice: only use this on software, sites and systems you own or are authorized to test. Disclaimer: Third-party repos we don't own or vouch for. AI-generated use cases can be wrong: review everything and run at your own risk. Full text: https://dedupedai.com/disclaimer/ URL: https://dedupedai.com/use-cases/mrpulor-gh-nuphus-mcp/drive-anti-detect-browser-over-cdp/ ## How to operate a native desktop app by its accessibility tree with Nuphus MCP Repo: [mrpulor-gh/nuphus-mcp](https://github.com/mrpulor-gh/nuphus-mcp) | Difficulty: Advanced | Time: 60 minutes | Basis: From the README Drive native UI through Windows UIA or macOS AX using observe, candidate, execute, and verify steps. **Who it is for:** An IT or operations person who needs an agent to use a desktop application that has no API or browser version. **The problem:** Pixel-coordinate automation breaks when window layout or scaling changes. **The outcome:** The agent selects UI elements by stable semantic locators and checks the resulting state after each action. ### Steps 1. Call `desktop_targets_list` to see the windows available to bind. 2. Call `desktop_target_bind` on the window you want the agent to control. 3. Call `desktop_semantic_observe`, then `desktop_semantic_candidate` to pick the element. 4. Call `desktop_semantic_execute` or `desktop_semantic_action`, then `desktop_verify_state` to confirm the result. ### Prompt to paste into your agent ```text Use the nuphus-mcp semantic desktop tools to bind the window titled , pick the element labeled , and perform . Return the verified state as text. Verify the state with desktop_verify_state before reporting success. ``` ### Caveat Semantic control is backed by Windows UIA and the macOS Accessibility API, and macOS needs the Accessibility permission for mouse and keyboard tools. Dual-use notice: only use this on software, sites and systems you own or are authorized to test. Disclaimer: Third-party repos we don't own or vouch for. AI-generated use cases can be wrong: review everything and run at your own risk. Full text: https://dedupedai.com/disclaimer/ URL: https://dedupedai.com/use-cases/mrpulor-gh-nuphus-mcp/operate-native-app-by-accessibility-tree/ ## How to combine a vision model with local OCR for exact clicks with Nuphus MCP Repo: [mrpulor-gh/nuphus-mcp](https://github.com/mrpulor-gh/nuphus-mcp) | Difficulty: Intermediate | Time: 45 minutes | Basis: From the README Use your own vision model for meaning and local OCR for pixel-precise coordinates on screen. **Who it is for:** A developer or analyst whose agent must find a labeled control on screen and click it reliably. **The problem:** A vision model can describe a screen but often cannot give exact coordinates for a click. **The outcome:** The agent gets a semantic description from the vision model and exact coordinates from OCR, then clicks and checks the result. ### Steps 1. Set `NUPHUS_MCP_VISION_API_KEY` and `NUPHUS_MCP_VISION_MODEL` in the MCP client env, and `NUPHUS_MCP_VISION_BASE_URL` if you use a non-default endpoint. 2. Call `desktop_vision` with a screenshot and a question about the screen. 3. Call `desktop_perceive` to get local OCR text and coordinates; models download on first run. 4. Use the returned coordinates with the mouse click tool, then take a screenshot to verify. ### Prompt to paste into your agent ```text Use the nuphus-mcp desktop_vision tool to describe where the button is on screen, then use desktop_perceive to find its exact coordinates. Click it and return a screenshot path. Verify the screenshot shows the expected change before you report success. ``` ### Caveat desktop_vision returns a clear error when it is not configured, and the README notes Zhipu GLM-4V-Flash caps output at 1024 tokens, so text-heavy screens may need a higher NUPHUS_MCP_VISION_MAX_TOKENS. Risk, credentials: Screenshots sent to your vision model provider can contain passwords or private data, so capture only the windows you intend to share. Dual-use notice: only use this on software, sites and systems you own or are authorized to test. Disclaimer: Third-party repos we don't own or vouch for. AI-generated use cases can be wrong: review everything and run at your own risk. Full text: https://dedupedai.com/disclaimer/ URL: https://dedupedai.com/use-cases/mrpulor-gh-nuphus-mcp/vision-ocr-precise-desktop-clicks/