DEDUPED Hot AI repos, turned into use cases Get 5 a day

How to combine a vision model with local OCR for exact clicks with Nuphus MCP

Use your own vision model for meaning and local OCR for pixel-precise coordinates on screen.

Authorization notice. Nuphus MCP is dual-use. Only use it on software, sites and systems you own or have explicit permission to test.

Credentials. Screenshots sent to your vision model provider can contain passwords or private data, so capture only the windows you intend to share.

Who it is forA developer or analyst whose agent must find a labeled control on screen and click it reliably.

The problemA vision model can describe a screen but often cannot give exact coordinates for a click.

The outcomeThe agent gets a semantic description from the vision model and exact coordinates from OCR, then clicks and checks the result.

What you need first

  • Rust stable toolchain to build from source with Cargo
  • Chrome or Edge installed for browser tools
  • An MCP client that supports stdio servers
  • Windows for full desktop control; macOS needs the Accessibility permission
  • Your own vision model API key only if you use desktop_vision

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 for your agent

Use the nuphus-mcp desktop_vision tool to describe where the <target label> 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.

Share: X LinkedIn Hacker News Email Copy link

Disclaimer

Deduped lists open-source repositories that belong to other people. We do not own, maintain, audit, or vouch for any of them, and a repo appearing here is not an endorsement or a safety check. The use cases, steps, and prompts are written by AI and may be incomplete, outdated, or just wrong. Where a use case is marked "Seen in the wild", we link to public posts where someone described something similar, but we have not verified those posts and the use case itself is still our AI's own write-up. Use all of this at your own risk. Only run code, commands, or agents on systems and accounts you are authorized to use, and read the code and every command before you run it, especially anything that touches credentials, files, or the network. Nothing here is professional, security, legal, or financial advice. All trademarks and project names belong to their respective owners, and their use here does not imply affiliation. Content is provided as is, without warranties of any kind. Found a mistake, a safety problem, or something you want removed? Email [email protected] and we will look at it promptly. Full disclaimer page