<?xml version="1.0" encoding="UTF-8"?><rss xmlns:dc="http://purl.org/dc/elements/1.1/" xmlns:content="http://purl.org/rss/1.0/modules/content/" xmlns:atom="http://www.w3.org/2005/Atom" version="2.0" xmlns:cc="http://cyber.law.harvard.edu/rss/creativeCommonsRssModule.html">
    <channel>
        <title><![CDATA[Stories by Sakshi Arora on Medium]]></title>
        <description><![CDATA[Stories by Sakshi Arora on Medium]]></description>
        <link>https://medium.com/@sakshiaroraresearch?source=rss-e2c037738a4b------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/1*fg_lFp-DsCGC-0DIGXog1w.jpeg</url>
            <title>Stories by Sakshi Arora on Medium</title>
            <link>https://medium.com/@sakshiaroraresearch?source=rss-e2c037738a4b------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Tue, 17 Nov 2026 20:49:06 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@sakshiaroraresearch/feed" rel="self" type="application/rss+xml"/>
        <webMaster><![CDATA[yourfriends@medium.com]]></webMaster>
        <atom:link href="http://medium.superfeedr.com/" rel="hub"/>
        <item>
            <title><![CDATA[Cursor’s 40-Tool Barrier: Navigating MCP Limits]]></title>
            <link>https://medium.com/@sakshiaroraresearch/cursors-40-tool-tango-navigating-mcp-limits-213a111dc218?source=rss-e2c037738a4b------2</link>
            <guid isPermaLink="false">https://medium.com/p/213a111dc218</guid>
            <dc:creator><![CDATA[Sakshi Arora]]></dc:creator>
            <pubDate>Mon, 14 Apr 2025 11:45:17 GMT</pubDate>
            <atom:updated>2025-06-09T09:32:49.995Z</atom:updated>
            <content:encoded><![CDATA[<p>Cursor’s Model Context Protocol (MCP) is a powerful feature that allows the AI code editor to extend its capabilities far beyond basic text manipulation. Think of it as a bridge that connects Cursor to a vast ecosystem of external tools and data sources. Instead of being confined to the information within your project files, Cursor, through MCP, can interact with databases, APIs, version control systems like GitHub, and a whole range of other applications. This enables advanced functionalities like:</p><ul><li><strong>Real-time data retrieval:</strong> Accessing live information from external sources.</li><li><strong>Automated workflows:</strong> Triggering actions in other systems based on your code.</li><li><strong>Enhanced code understanding:</strong> Leveraging external documentation and libraries.</li></ul><p>Essentially, MCP transforms Cursor from a smart editor into a dynamic, interconnected development environment.</p><h3><strong>UPDATE: With Cursor version 50.03 onwards, just clicking on tool, will disable a particular tool.</strong></h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*Fk5o9rotwZ8D9W5BhA0LAQ.png" /></figure><p>Cursor’s Model Context Protocol (MCP) currently has limitations regarding the number of tools it can effectively handle. Specifically, there’s a practical limit around 40 tools. This constraint has generated discussions within the Cursor community, with users seeking ways to overcome it. Here’s a breakdown of the issue and potential workarounds:</p><h3><strong>Understanding the MCP Tool Limit</strong></h3><p><strong>MCP’s Purpose:<br></strong>The MCP is designed to extend Cursor’s capabilities by integrating external tools and data sources. This allows the AI to interact with various systems, like GitHub, databases, and other applications.</p><p><strong>The Limitation:<br></strong>Due to the complexities of managing and selecting from a large number of tools, Cursor currently performs optimally when the number of active MCP tools is kept within a certain range, approximately 40. Exceeding this limit can lead to performance issues and reduced accuracy in tool selection.</p><p><strong>Reasons for the Limit:<br></strong>Managing a vast array of tools poses challenges for the AI in accurately determining which tool to use in a given context. Context window limitations within current LLMs can also be a factor.</p><h3><strong>Workarounds and Solutions</strong></h3><p>Based on community discussions and available information, here are some strategies to mitigate the 40-tool limit:</p><p><strong>Selective Tool Activation:<br></strong>The most immediate workaround is to carefully select and activate only the most essential MCP tools. By disabling unnecessary tools, users can stay within the limit and ensure that the AI focuses on the most relevant functionalities.</p><p><strong>MCP Tool Management:<br></strong>There are community driven efforts to create tools that help manage the MCP tools. One example is a MCP server hub that can manage other MCP servers. This allows you to have many MCP servers, but only use a few tools that manage those servers. Basically idea is that “MCP server hubs” consolidates multiple MCP servers into a smaller set of tools, effectively bypassing the limit.</p><p><strong>Feature Requests and Future Development:</strong></p><p>The Cursor development team is aware of this limitation, and there are ongoing discussions about potential solutions, including:</p><ul><li>Improved tool selection algorithms.</li><li>Increase MCP tool limit <a href="https://forum.cursor.com/t/increase-the-mcp-tool/69194">https://forum.cursor.com/t/increase-the-mcp-tool/69194</a></li><li>Mechanisms for filtering MCP tools. <a href="https://forum.cursor.com/t/add-the-possibility-to-filter-mcp-tools/76776">https://forum.cursor.com/t/add-the-possibility-to-filter-mcp-tools/76776</a></li><li>Refinements to how cursor manages the context windows.</li></ul><p><strong>Key Takeaways</strong></p><ul><li>The 40-tool limit in Cursor’s MCP is a current constraint that users should be aware of.</li><li>Strategic tool management and community-developed solutions can help mitigate this limitation.</li><li>The Cursor team is actively working on improvements to address this issue.</li></ul><p>It’s important to keep an eye on Cursor’s official documentation and community forums for the latest updates and solutions regarding MCP tool management.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=213a111dc218" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[A picture is worth a thousand words..]]></title>
            <link>https://medium.com/@sakshiaroraresearch/a-picture-is-worth-a-thousand-words-78c2a45b88be?source=rss-e2c037738a4b------2</link>
            <guid isPermaLink="false">https://medium.com/p/78c2a45b88be</guid>
            <dc:creator><![CDATA[Sakshi Arora]]></dc:creator>
            <pubDate>Wed, 11 Dec 2024 10:31:45 GMT</pubDate>
            <atom:updated>2024-12-11T10:31:45.057Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*z8j7BhtlZ2zg5EEI.jpeg" /></figure><p>… we have all heard this. Our love for comic books and reactions to visuals has scientifically proven it.</p><blockquote><em>visual stimulation over text translation allows the brain to consume the material with more consummate ease — </em><a href="https://www.easel.ly/blog/text-vs-images-which-content-format-effective/"><em>www.easel.ly</em></a></blockquote><p>Let’s use this insight to collect data easily, playfully and in a much more engaging fashion, reducing the form abandonment rate.</p><blockquote>The human brain can process entire images that the eye sees for as little as 13 milliseconds. — <a href="http://news.mit.edu/2014/in-the-blink-of-an-eye-0116">MIT research</a></blockquote><p>Visual content identification make data collection intuitive and language agnostic — no need to implement costly language translators ($$$) and no need to read to understand an option. Additionally this gives a less error prone approach for form filler filling 100&#39;s of fields. Visually capturing data can drastically reduce the data collection errors.</p><blockquote>People following directions with text and illustrations do 323 percent better than those following directions without illustrations. — <a href="https://scholar.google.co.uk/scholar?as_q=&amp;num=10&amp;btnG=Search+Scholar&amp;as_epq=&amp;as_oq=&amp;as_eq=&amp;as_occt=any&amp;as_sauthors=%22W.%20Howard%20Levie%22&amp;as_publication=&amp;as_ylo=&amp;as_yhi=&amp;as_allsubj=all&amp;hl=en">W. Howard Levie</a> &amp; <a href="https://scholar.google.co.uk/scholar?as_q=&amp;num=10&amp;btnG=Search+Scholar&amp;as_epq=&amp;as_oq=&amp;as_eq=&amp;as_occt=any&amp;as_sauthors=%22Richard%20Lentz%22&amp;as_publication=&amp;as_ylo=&amp;as_yhi=&amp;as_allsubj=all&amp;hl=en">Richard Lentz</a> (<a href="https://link.springer.com/article/10.1007%2FBF02765184">Article</a>)</blockquote><p>AEM Forms allows you to collect data using an image, a 3D graphic, or a video without overwhelming a user.</p><p>Some use cases are:</p><h3>Basic Details: State selection from a map, to capture address.</h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*8ETOcVUBogHwwnqe.png" /></figure><h3>Commerce: Lifestyle store images, to indicate which all products in the image one intends to buy.</h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*GIPyg5p8_O4xspqm.jpeg" /></figure><h3>Medical Details: Teeth selection in the jaw image to indicate which tooth needs treatment</h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*YVDXbSQb-uTGLjJM.jpeg" /></figure><h3>Insurance: Damage capture on car.</h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*irIPp-uPmE-9heyB.jpeg" /></figure><p>and so on …</p><p>Lets take an example of a Kitchen configurator, where a buyer can choose multiple products just by tapping various items in a kitchen image.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/630/0*buT69W2bxHqgpxAW.gif" /></figure><p>You can see, when a form filler selects an item in the image, the corresponding checkbox is automatically selected. Conversely, selecting or unselecting the checkbox also selects or unselects the corresponding portion of the image.</p><p>To replicate similar experience in your form:</p><ol><li>Create a <a href="https://github.com/adobe/aem-core-forms-components/tree/master/it/apps/src/main/content/jcr_root/apps/forms-core-components-it/componentDef/svg">custom component</a> to upload the SVG Image to form.</li><li>Create a <a href="https://github.com/adobe/aem-core-forms-components/tree/master/it/apps/src/main/content/jcr_root/apps/forms-core-components-it/componentDef/svgSelector">composite component</a> by extending the forms panel component. It would contain two panels, one containing image and another would contain a set of corresponding checkboxes.</li><li>Create proxy components for <a href="https://github.com/adobe/aem-core-forms-components/tree/master/it/apps/src/main/content/jcr_root/apps/forms-core-components-it/components/svg">custom SVG component</a> and <a href="https://github.com/adobe/aem-core-forms-components/tree/master/it/apps/src/main/content/jcr_root/apps/forms-core-components-it/components/svgSelector">composite component</a>.</li><li>Create a client library. Use these sample <a href="https://github.com/adobe/aem-core-forms-components/blob/master/it/apps/src/main/content/jcr_root/apps/forms-core-components-it/clientlibs/svg-selector/js/functions.js">JS</a> and <a href="https://github.com/adobe/aem-core-forms-components/blob/master/it/apps/src/main/content/jcr_root/apps/forms-core-components-it/clientlibs/svg-selector/styles/site.css">CSS</a> files. Add the client library to the <a href="https://github.com/adobe/aem-core-forms-components/blob/master/it/content/src/main/content/jcr_root/conf/core-components-it/settings/wcm/policies/.content.xml#L468">template page</a>.</li><li>Create a form, and use the custom SVG component and composite component to the form.</li><li>Add the required checkboxes. The field names should match the attribute of the uploaded SVG image. You can find a sample form <a href="https://github.com/adobe/aem-core-forms-components/tree/master/it/content/src/main/content/jcr_root/content/forms/af/core-components-it/samples/testsvg">here</a>.</li></ol><p>Voila! your customers can now simply tap parts of the image to select objects of their choice. Try it and measure improvements in conversion rate and reduction in Abandonment rate. Happy form creating!</p><p>Feel free to reach out to me at:<br><a href="https://www.linkedin.com/in/sakshiarora4/">https://www.linkedin.com/in/sakshiarora4/</a></p><p><strong>References</strong></p><p><a href="https://www.learnevents.com/learning-insights/imagery-vs-text-which-does-the-brain-prefer/#:~:text=According%20to%20Zabisco%2C%20the%20average,material%20with%20more%20consummate%20ease">https://www.learnevents.com/learning-insights/imagery-vs-text-which-does-the-brain-prefer/#:~:text=According%20to%20Zabisco%2C%20the%20average,material%20with%20more%20consummate%20ease</a></p><p><a href="https://www.easel.ly/blog/text-vs-images-which-content-format-effective/">https://www.easel.ly/blog/text-vs-images-which-content-format-effective/</a></p><p><a href="https://journal.lib.uoguelph.ca/index.php/perj/article/view/3137">https://journal.lib.uoguelph.ca/index.php/perj/article/view/3137</a></p><p><a href="https://oit.williams.edu/files/2010/02/using-images-effectively.pdf">https://oit.williams.edu/files/2010/02/using-images-effectively.pdf</a></p><p><a href="https://www.mdpi.com/2504-3900/1/9/959">https://www.mdpi.com/2504-3900/1/9/959</a></p><p><a href="https://www.nngroup.com/articles/picture-superiority-effect/">https://www.nngroup.com/articles/picture-superiority-effect/</a></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=78c2a45b88be" width="1" height="1" alt="">]]></content:encoded>
        </item>
    </channel>
</rss>