<?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 Isitaghosh on Medium]]></title>
        <description><![CDATA[Stories by Isitaghosh on Medium]]></description>
        <link>https://medium.com/@isitaghosh28?source=rss-8b4b40a707e0------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/1*MqTjn5tvxuHW4X6OfUk-qQ@2x.jpeg</url>
            <title>Stories by Isitaghosh on Medium</title>
            <link>https://medium.com/@isitaghosh28?source=rss-8b4b40a707e0------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Thu, 19 Nov 2026 04:45:21 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@isitaghosh28/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[What Happens When You Change Your Delivery Address After Placing an Order?]]></title>
            <link>https://medium.com/@isitaghosh28/what-happens-when-you-change-your-delivery-address-after-placing-an-order-c62b0ffa9a2b?source=rss-8b4b40a707e0------2</link>
            <guid isPermaLink="false">https://medium.com/p/c62b0ffa9a2b</guid>
            <category><![CDATA[technology]]></category>
            <category><![CDATA[ecommerce]]></category>
            <category><![CDATA[supply-chain]]></category>
            <category><![CDATA[api-integration]]></category>
            <category><![CDATA[mulesoft-integration]]></category>
            <dc:creator><![CDATA[Isitaghosh]]></dc:creator>
            <pubDate>Wed, 18 Nov 2026 04:31:03 GMT</pubDate>
            <atom:updated>2026-11-18T04:31:03.830Z</atom:updated>
            <content:encoded><![CDATA[<p>You order a pair of headphones for home. Ten minutes later, you remember: tomorrow is an office day. You open the app, change the address, and tap Save. 📦</p><p>Done? On your screen, perhaps. Behind the scenes, a few more conversations need to happen.</p><figure><img alt="An address change shared between an online store, warehouse and delivery partner through MuleSoft." src="https://cdn-images-1.medium.com/proxy/1*PXqC5QdAo7FrhQrsfNGz7A.png" /></figure><h3>Who needs to hear about the new address?</h3><p>The store updates your order. The warehouse may need to replace the shipping label. If the parcel has left, the delivery partner needs to accept the new destination and adjust its delivery plan.</p><p>Changing the address saved in your profile may not update an existing order. It needs to be an address change for that particular order.</p><h3>Can I change it after dispatch?</h3><p>Sometimes. A parcel waiting for collection is easier to redirect than one already on the driver’s route. The retailer’s rules, delivery area and courier’s capabilities decide what is possible. A new destination may also mean a delay or fee.</p><h3>Where does MuleSoft fit in?</h3><p>Think of it as a coordinator between systems that don’t naturally speak the same language. In a setup using MuleSoft, it can pass the address update from the store to the warehouse and courier, translate their data formats, and bring their replies back.</p><p>These connections are often called APIs: agreed ways for software systems to exchange information.</p><h3>What does the update actually look like?</h3><p>Here’s a made-up request payload — the small bundle of information a system sends:</p><pre>{<br>  &quot;requestId&quot;: &quot;CHANGE-104&quot;,<br>  &quot;orderId&quot;: &quot;ORD-4821&quot;,<br>  &quot;newDeliveryAddress&quot;: {<br>    &quot;street&quot;: &quot;12 Example Lane&quot;,<br>    &quot;city&quot;: &quot;Kolkata&quot;,<br>    &quot;postalCode&quot;: &quot;700029&quot;<br>  }<br>}</pre><p>The order ID identifies the order. The address says where it should go. The request ID helps track this particular change.</p><h3>What if the store says yes, but the courier doesn’t?</h3><p>Quick thought: the store displays “Address updated,” but the courier is offline. Would you assume the parcel is heading to the new address?</p><p>This is where a good design matters. It should track each reply and show <strong>pending</strong> until the required systems confirm. Temporary failures can be retried safely; an invalid address or rejected reroute needs correction or support. If only some systems accept, the process must resolve that mismatch and tell the customer what happened.</p><h3>Could someone change another person’s order?</h3><p>The service should verify who is making the request and whether they’re allowed to update that order. Knowing an order number alone should never be enough. The new address also needs checking for completeness and delivery coverage.</p><h3>What if I tap Save twice — or change my mind again?</h3><p>Two taps should not create two rerouting actions. A request ID can help the receiving system recognise a repeated request. Developers call this <em>idempotency</em>: repeating the same request has the same effect as sending it once.</p><p>A second, different address change needs its own identity and an order version or sequence number. That helps prevent a delayed older update from overwriting the latest one.</p><h3>Does MuleSoft guarantee delivery to the new address?</h3><p>No. It helps coordinate information, but the courier must accept and carry out the change. Checks, retries and customer notifications have to be designed into the process; installing MuleSoft doesn’t provide them automatically.</p><h3>Why not connect the systems directly?</h3><p>You can. For a small setup, direct connections may be enough. MuleSoft can be useful as more systems join, because teams can reuse connections and manage data conversion, security and failures in a consistent way.</p><p>That little Save button starts a chain of handoffs. The reassuring part is knowing which of them have actually finished. 💡</p><p><strong>If you were designing this app, when would you show “Address updated”: after the store saves it, or after the courier confirms it — and why?</strong></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=c62b0ffa9a2b" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Your API Timed Out. The Payment Went Through. So Why Was the Customer Charged Twice?]]></title>
            <link>https://medium.com/@isitaghosh28/your-api-timed-out-the-payment-went-through-so-why-was-the-customer-charged-twice-683384781f8a?source=rss-8b4b40a707e0------2</link>
            <guid isPermaLink="false">https://medium.com/p/683384781f8a</guid>
            <category><![CDATA[microservices]]></category>
            <category><![CDATA[distributed-systems]]></category>
            <category><![CDATA[idempotency]]></category>
            <category><![CDATA[api-design]]></category>
            <category><![CDATA[backend-development]]></category>
            <dc:creator><![CDATA[Isitaghosh]]></dc:creator>
            <pubDate>Wed, 04 Nov 2026 04:31:04 GMT</pubDate>
            <atom:updated>2026-11-04T04:31:04.117Z</atom:updated>
            <content:encoded><![CDATA[<h4>One timeout. One retry. Two successful payments. This is where idempotency becomes important.</h4><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*b7hncP6NC3u2kJsPpXWwGw.png" /></figure><p>Imagine this.</p><p>A customer is paying ₹1,000 through your application. They click Pay Now, but the page keeps loading.</p><p>After a few seconds, they see a timeout message.</p><p>Naturally, they click Pay Now again.</p><p>This time, everything works.</p><p>But there’s a problem: the first payment was already successful.</p><p>The customer has now been charged ₹2,000 instead of ₹1,000.</p><p>So, what went wrong?</p><h3>A timeout doesn’t always mean failure</h3><p>When an API request times out, it doesn’t necessarily mean the operation failed.</p><p>Sometimes:</p><ul><li>The server processed the request successfully.</li><li>The response got delayed or lost.</li><li>The client never received confirmation.</li><li>The client retried the request, creating another transaction.</li></ul><p>This is especially risky in payment, order, booking, and fund transfer systems.</p><p>And simply retrying the request isn’t always the solution.</p><h3>What is Idempotency?</h3><p>This is where idempotency comes in.</p><p>In simple terms, idempotency ensures that sending the same request multiple times produces the same business effect as sending it once.</p><p>For example, if a payment request is submitted twice with the same idempotency key, the system should process the payment only once and return the original result for the retry.</p><p>A request might look like this:</p><pre>POST /payments<br>Idempotency-Key: payment-12345</pre><p>The idea is simple:</p><ol><li>Generate a unique key for the business operation.</li><li>Send that key with the API request.</li><li>Before processing, check whether the key has already been used.</li><li>If it has, return the previously stored result.</li><li>Otherwise, process the request and store the result against that key.</li></ol><p>Now, even if the client retries, the payment won’t be processed twice.</p><h3>A few things developers should watch out for</h3><p>Implementing idempotency isn’t just about adding a header.</p><ul><li>Don’t generate a new key for every retry. The same business operation must reuse the same key.</li><li>Handle simultaneous requests. Two identical requests can arrive at almost the same time. Use atomic database operations or unique constraints to prevent both from being processed.</li><li>Validate the request. Reusing the same key with a different payload should not silently return an unrelated result.</li><li>Think about key expiry. Decide how long to retain keys based on your business requirements.</li></ul><h3>Where is it useful?</h3><p>Idempotency is particularly important in:</p><ul><li>Payment processing</li><li>Order creation</li><li>Flight or hotel bookings</li><li>Fund transfers</li><li>Subscription management</li></ul><p>One important distinction: HTTP methods like GET and PUT are defined as idempotent in terms of their intended effects. POST is not inherently idempotent, but we can make a POST operation idempotent through application-level design.</p><h3>One final thought</h3><p>Retries are necessary in distributed systems. Networks fail, responses get lost, and services become temporarily unavailable.</p><p>But a reliable API isn’t just one that can retry a request.</p><p>It’s one that knows how to handle the retry safely.</p><p>Have you ever faced duplicate records or transactions because of an API retry? How did you handle it?</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=683384781f8a" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Your API Returned 200. So Why Did the Business Process Fail?]]></title>
            <link>https://medium.com/@isitaghosh28/your-api-returned-200-so-why-did-the-business-process-fail-ebe40d4285b8?source=rss-8b4b40a707e0------2</link>
            <guid isPermaLink="false">https://medium.com/p/ebe40d4285b8</guid>
            <dc:creator><![CDATA[Isitaghosh]]></dc:creator>
            <pubDate>Tue, 06 Oct 2026 04:31:02 GMT</pubDate>
            <atom:updated>2026-10-06T04:31:02.979Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*7q9Cf0Wo6kreIRN-DrkdVA.png" /></figure><p>Have you ever seen this?</p><p>The API returned <strong>200 OK</strong> ✅<br>No exception.<br>No timeout.<br>Everything looked fine in the logs.</p><p>And then someone says:</p><p><strong>“The transaction was not completed.”</strong> 😐</p><p>Wait… but the API returned 200?</p><p>This is something I’ve come across while working on integrations, and it made me realize that <strong>a successful API response doesn’t always mean a successful business process.</strong></p><h3>🤔 What does 200 OK actually tell us?</h3><p>Usually, 200 OK means the request was successfully processed by the server.</p><p>But it doesn’t necessarily mean that <strong>everything that happened after that was successful.</strong></p><p>For example, imagine an order process:</p><pre>Customer<br>   ↓<br>Order API<br>   ↓<br>Payment<br>   ↓<br>Inventory<br>   ↓<br>Shipping</pre><p>The Order API might return:</p><pre>200 OK ✅</pre><p>But what if the payment was successful and the order was created, but the inventory update failed?</p><p>Technically, the API call succeeded.</p><p>But from a business point of view…</p><p><strong>The order process didn’t fully succeed.</strong> 😅</p><h3>⚠️ Sometimes the problem is even harder to find</h3><p>Imagine the API returns:</p><pre>{<br>  &quot;status&quot;: &quot;SUCCESS&quot;,<br>  &quot;amount&quot;: 5000<br>}</pre><p>HTTP status?</p><p><strong>200 OK ✅</strong></p><p>The response looks perfectly fine.</p><p>But what if the amount is wrong?</p><p>There may be no exception.</p><p>There may be no timeout.</p><p>There may be nothing obvious in the logs.</p><p>But the business result is still <strong>wrong</strong>.</p><p>And honestly, these are sometimes more difficult to find than a simple API failure.</p><h3>🔗 This is why integration can get tricky</h3><p>In an integration, one request can involve multiple systems.</p><p>For example:</p><pre>API<br> ↓<br>Validate data<br> ↓<br>Create record<br> ↓<br>Update database<br> ↓<br>Call another system<br> ↓<br>Send notification</pre><p>Now imagine one of the middle steps fails.</p><p>What should happen?</p><p>Should we retry?</p><p>Should we rollback?</p><p>Should we continue?</p><p>Should we mark the transaction as partially completed?</p><p>These are not just technical questions.</p><p><strong>They are business decisions too.</strong></p><h3>🔧 So how do we troubleshoot it?</h3><p>When the API says 200 OK but the business process failed, I wouldn&#39;t immediately start changing the API code.</p><p>I would go step by step.</p><h3>1️⃣ Check the complete transaction</h3><p>Don’t stop at the API response.</p><p>Check what happened to the transaction after the API returned 200.</p><pre>API → Payment → Inventory → Shipping → Notification</pre><p>Find the exact point where the process stopped.</p><h3>2️⃣ Use a correlation or transaction ID</h3><p>Make sure the same transaction ID is available across the systems involved.</p><p>For example:</p><pre>Transaction ID: ORD-12345</pre><p>Now search for ORD-12345 in the logs of each system.</p><p>This makes it much easier to answer:</p><p><strong>“Where did this transaction actually fail?”</strong></p><h3>3️⃣ Check the downstream response</h3><p>Don’t assume that because your API returned 200, every downstream system succeeded.</p><p>Check:</p><ul><li>Response status</li><li>Response body</li><li>Error message</li><li>Business status</li><li>Timeout</li><li>Validation failure</li></ul><p>Sometimes the downstream system doesn’t throw an obvious technical error, but the <strong>business operation still fails.</strong></p><h3>4️⃣ Check for partial processing</h3><p>This is an important one.</p><p>Maybe:</p><pre>Order created       ✅<br>Payment completed   ✅<br>Inventory updated   ❌<br>Shipping            ❌</pre><p>Now you have a <strong>partially completed transaction.</strong></p><p>Before retrying the whole request, understand what has already happened.</p><p>Otherwise, you might create duplicate orders or duplicate payments. 😬</p><h3>5️⃣ Check whether it’s safe to retry</h3><p>Ask yourself:</p><blockquote><strong><em>“If I send the same request again, will I create something twice?”</em></strong></blockquote><p>This is where <strong>idempotency</strong> becomes important.</p><p>For example, if the first request already created an order but failed while updating inventory, simply retrying the entire request could create another order.</p><p>Depending on the design, it may be safer to retry only the failed operation.</p><h3>6️⃣ Validate the actual business result</h3><p>Finally, don’t validate only this:</p><pre>HTTP 200 ✅</pre><p>Also check:</p><pre>Order created?       ✅<br>Payment completed?   ✅<br>Inventory updated?   ❌<br>Shipping started?    ❌</pre><p><strong>The final business state is what really matters.</strong></p><h3>💡 A simple approach I like</h3><p>When debugging these kinds of issues, I usually think about three questions:</p><p><strong>1. Did the API request succeed?</strong></p><p><strong>2. Did all the downstream systems succeed?</strong></p><p><strong>3. Did we get the business outcome we actually wanted?</strong></p><p>If the answer to #1 is <strong>yes</strong>, but #2 or #3 is <strong>no</strong>, then 200 OK doesn&#39;t tell the whole story.</p><h3>💭 One thing I’ve learned</h3><p><strong>200 OK doesn’t always mean “everything is successful&quot;.</strong></p><p>Sometimes it only means:</p><blockquote><em>“This particular request was handled successfully.”</em></blockquote><p>The actual business process might still be:</p><p>❌ Failed<br>⚠️ Partially completed<br>❌ Incorrect</p><p>That’s why I feel integration is not just about connecting <strong>System A to System B</strong>.</p><p>It’s also about making sure the <strong>actual business outcome is correct.</strong></p><p>And this applies whether you’re working with <strong>MuleSoft, Java, Node.js, APIs, or any other integration/backend technology.</strong></p><h3>💬 What do you think?</h3><p>If an API returns <strong>200 OK</strong>, but the actual business transaction wasn’t completed, would you consider it a successful API call?</p><p><strong>How do you usually troubleshoot this kind of issue in your projects?</strong> 👇</p><p>#APIs #Integration #SoftwareEngineering #MuleSoft #BackendDevelopment #DevOps</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=ebe40d4285b8" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[ 5 DataWeave Mistakes MuleSoft Developers Make (and How to Fix Them)]]></title>
            <link>https://medium.com/@isitaghosh28/5-dataweave-mistakes-mulesoft-developers-make-and-how-to-fix-them-79c725346871?source=rss-8b4b40a707e0------2</link>
            <guid isPermaLink="false">https://medium.com/p/79c725346871</guid>
            <category><![CDATA[dataweave]]></category>
            <category><![CDATA[software-development]]></category>
            <category><![CDATA[api-development]]></category>
            <category><![CDATA[programming]]></category>
            <category><![CDATA[mulesoft]]></category>
            <dc:creator><![CDATA[Isitaghosh]]></dc:creator>
            <pubDate>Tue, 29 Sep 2026 04:31:03 GMT</pubDate>
            <atom:updated>2026-09-29T04:31:03.845Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*iJ_XXwKHzFnfUSaNy6POGg.png" /></figure><p>DataWeave is one of the most useful parts of MuleSoft. It helps us transform data from one format into another.</p><p>It can look easy at first. But once you start working with real-world data, things can get tricky. 😅</p><p>Here are 5 common mistakes that can cause problems—especially when your application moves from testing to production.</p><h3>1️⃣ Using map when you actually need mapObject</h3><p>Let’s start with a simple one.</p><p>Imagine you have a list of users:</p><pre>[<br>  {&quot;name&quot;: &quot;John&quot;},<br>  {&quot;name&quot;: &quot;Sarah&quot;}<br>]</pre><p>Because this is a <strong>list</strong>, you can use map to go through each user:</p><pre>%dw 2.0<br>output application/json<br>---<br>payload map (user) -&gt; user.name</pre><p>Result:</p><pre>[<br>  &quot;John&quot;,<br>  &quot;Sarah&quot;<br>]</pre><p>But what if you’re working with an <strong>object</strong> instead of a list?</p><p>That’s where mapObject comes in.</p><p>💡 <strong>Simple rule:</strong></p><p><strong>List/Array → </strong><strong>map</strong></p><p><strong>Object → </strong><strong>mapObject</strong></p><p>If you’re not sure which one to use, first check what type of data you’re working with.</p><h3>2️⃣ Forgetting that a value can be null</h3><p>This is one of those problems that often doesn’t show up during testing. 😅</p><p>Imagine you want to combine a person’s first and last name:</p><pre>payload.firstName ++ &quot; &quot; ++ payload.lastName</pre><p>Your test data looks like this:</p><pre>{<br>  &quot;firstName&quot;: &quot;John&quot;,<br>  &quot;lastName&quot;: &quot;Doe&quot;<br>}</pre><p>So everything works:</p><pre>John Doe</pre><p>👍 Great!</p><p>But production sends:</p><pre>{<br>  &quot;firstName&quot;: &quot;John&quot;,<br>  &quot;lastName&quot;: null<br>}</pre><p>Now you’re trying to combine text with a null value.</p><p>Your transformation 💥</p><p>You can provide a backup value:</p><pre>(payload.firstName default &quot;&quot;) ++ &quot; &quot; ++ (payload.lastName default &quot;&quot;)</pre><p>Now, if lastName is null, DataWeave uses an empty string instead.</p><p>This isn’t only about joining names.</p><p>The same problem can happen when:</p><p>🔹 Calling a function<br>🔹 Doing calculations<br>🔹 Reading an API response<br>🔹 Working with optional fields</p><p>💡 <strong>Always ask: “What happens if this value is null?”</strong></p><p>Real-world data isn’t always as clean as test data. 😄</p><h3>3️⃣ Not using update when you only need to change one field</h3><p>Sometimes developers rebuild an entire object just to change one value.</p><p>For example:</p><pre>{<br>  &quot;name&quot;: &quot;John&quot;,<br>  &quot;age&quot;: 30,<br>  &quot;city&quot;: &quot;Kolkata&quot;<br>}</pre><p>If you only want to change age, DataWeave&#39;s update operator can make this much cleaner:</p><pre>payload update {<br>    case age at .age -&gt; 31<br>}</pre><p>Instead of manually recreating the whole object.</p><p>💡 <strong>If you’re only changing one or two fields, don’t rebuild the entire payload unnecessarily.</strong></p><h3>4️⃣ Forgetting about performance when the data gets bigger</h3><p>A transformation might work perfectly with 10 records.</p><p>Then someone gives it 100,000 records. 😳</p><p>Suddenly, the same transformation takes much longer.</p><p>For example, imagine you have a list of orders and a list of customers.</p><p>For every order, you search through the entire customer list to find the matching customer:</p><pre>payload.orders map (order) -&gt;<br>    payload.customers filter (customer) -&gt;<br>        customer.id == order.customerId</pre><p>This may work with a small amount of data.</p><p>But with large datasets, repeatedly searching through the same list can become expensive.</p><p>💡 <strong>Don’t only test whether your transformation works. Think about how it will behave with real production data.</strong></p><p>Ask yourself:</p><p><strong>“What happens if this payload becomes 10x or 100x bigger?”</strong></p><h3>5️⃣ Making DataWeave more complicated than it needs to be</h3><p>This one is very easy to do. 😂</p><p>Sometimes we see a requirement and think:</p><blockquote><em>“Let’s create five variables, three filters, two maps, a reduce… and maybe one more function just for fun.” 😅</em></blockquote><p>But sometimes the solution can be much simpler.</p><p>For example:</p><pre>%dw 2.0<br>output application/json<br>---<br>payload.users<br>    filter ((user) -&gt; user.active)<br>    map ((user) -&gt; user.name)</pre><p>This simply:</p><ol><li>Takes the users</li><li>Keeps the active users</li><li>Gets their names</li></ol><p>That’s it.</p><p><strong>Simple code is often easier to understand, test, debug, and maintain.</strong></p><p>You don’t always need the cleverest DataWeave expression.</p><h3>🚀 The Takeaway</h3><p>When writing DataWeave, don’t only ask:</p><p><strong>“Does my transformation work?”</strong></p><p>Also ask:</p><p>🔹 What type of data am I working with?</p><p>🔹 What happens if a value is null?</p><p>🔹 What happens when the data becomes much larger?</p><p>🔹 Can another developer easily understand this?</p><p>These small questions can make a big difference in production.</p><p><strong>Good DataWeave isn’t just about getting the right output. It’s about writing transformations that are reliable, readable, and ready for real-world data.</strong></p><h3>💬 Let’s discuss</h3><p>What’s the <strong>DataWeave problem that took you the longest to debug?</strong> 😄</p><p>Maybe your experience can save someone else a few hours of debugging. 👇</p><p>#MuleSoft #DataWeave #MuleSoftDeveloper #Integration #APIIntegration #DataTransformation</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=79c725346871" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[MuleSoft Is Moving Beyond APIs: What MCP Means for Integration Developers]]></title>
            <link>https://medium.com/@isitaghosh28/mulesoft-is-moving-beyond-apis-what-mcp-means-for-integration-developers-7453c5e92a8d?source=rss-8b4b40a707e0------2</link>
            <guid isPermaLink="false">https://medium.com/p/7453c5e92a8d</guid>
            <category><![CDATA[mulesoft-integration]]></category>
            <category><![CDATA[mcp-server]]></category>
            <category><![CDATA[mulesoft-ai-integration]]></category>
            <category><![CDATA[software-engineering]]></category>
            <category><![CDATA[mulesoft]]></category>
            <dc:creator><![CDATA[Isitaghosh]]></dc:creator>
            <pubDate>Fri, 25 Sep 2026 13:27:29 GMT</pubDate>
            <atom:updated>2026-09-25T13:27:29.110Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/768/1*S6KWeBiSgvsRv7GFFFyHjg@2x.jpeg" /></figure><p>MuleSoft developers are used to connecting APIs, transforming data with DataWeave, securing integrations, and working with enterprise systems.</p><p>But now there’s another consumer entering the picture:</p><p><strong>AI agents</strong>. 🤖</p><p>So the interesting question is:</p><p>💭 <strong>What changes when an AI agent becomes a consumer of your MuleSoft APIs?</strong></p><p>🔗 <strong>Where does MuleSoft fit?</strong></p><p>A traditional integration might look like:</p><pre>Application<br>    ↓<br>MuleSoft API<br>    ↓<br>Backend System</pre><p>With AI agents, we can have:</p><pre>AI Agent<br>    ↓<br>   MCP<br>    ↓<br>MuleSoft API<br>    ↓<br>Enterprise System</pre><p><strong>MuleSoft doesn’t suddenly become irrelevant.</strong></p><p>API security, DataWeave, policies, governance, error handling, and enterprise connectivity are still important.</p><p>The difference is the <strong>type of consumer.</strong></p><p><strong>A Practical Example</strong></p><p>Imagine an e-commerce company where order information is available through a MuleSoft API</p><pre>GET /orders/{orderId}</pre><p>A customer-service employee asks:</p><p>💬 “What’s the status of order 100245?”</p><p>The flow could be:</p><pre>   User<br>     ↓<br>AI Agent<br>     ↓<br>getOrderStatus<br>     ↓<br>MuleSoft API<br>     ↓<br>Order System</pre><p>The backend returns:</p><pre>{<br>  &quot;ORDER_ID&quot;: &quot;100245&quot;,<br>  &quot;STATUS&quot;: &quot;SHIPPED&quot;,<br>  &quot;DELIVERY_DATE&quot;: &quot;2026-09-28&quot;<br>}</pre><p>MuleSoft can use DataWeave to provide a clean response:</p><pre>%dw 2.0<br>output application/json<br>---<br>{<br>    orderId: payload.ORDER_ID,<br>    status: payload.STATUS,<br>    expectedDelivery: payload.DELIVERY_DATE<br>}</pre><p>The AI agent can then respond:</p><p>📦<strong> “Order 100245 has been shipped and is expected to arrive on September 28.”</strong></p><p>Simple, right?</p><p>But here’s where things get interesting. 👀</p><p>🔐 <strong>Should the agent access everything?</strong></p><p>Imagine the API also has:</p><pre>GET    /orders/{id}<br>PUT    /orders/{id}<br>DELETE /orders/{id}<br>POST   /orders/{id}/refund</pre><p>If the agent only needs to check an order status, <strong>why should it have permission to refund or delete an order?</strong></p><p>👉 <strong>Give the agent only the capabilities it actually needs.</strong></p><p>This is the same <strong>least-privilege principle</strong> we already apply to API security.</p><p>🧩 <strong>What does this mean for MuleSoft developers?</strong></p><p>You don’t need to throw away everything you already know. 😄</p><p>Your skills in:</p><p>🔹 API design</p><p>🔹 DataWeave</p><p>🔹 Security</p><p>🔹 Authentication &amp; authorization</p><p>🔹 API policies</p><p>🔹 Governance</p><p>🔹 Enterprise integration</p><p>are still highly relevant.</p><p>The question is simply changing from:</p><blockquote><strong>“How should an application consume my API?”</strong></blockquote><p>to:</p><blockquote><strong>“How should an AI agent safely consume this capability?”</strong></blockquote><p>🚀 The bigger picture</p><p>Traditional:</p><pre>Application → API → MuleSoft → Backend</pre><p>Increasingly:</p><pre>AI Agent → MCP → MuleSoft → Enterprise Systems</pre><p><strong>AI agents aren’t replacing integration. They’re becoming another consumer that integration developers need to design for.</strong></p><p>And that’s where our existing MuleSoft skills still have a role to play.</p><p>💬 <strong>Let’s discuss</strong></p><p>If you’re a MuleSoft developer:</p><p><strong>Would you expose an existing MuleSoft API as an AI capability, or would you add another control layer between the AI agent and the API?</strong></p><p>👇 I’d love to hear your thoughts.</p><p>#MuleSoft #MCP #AIAgents #Integration #APIs #DataWeave #EnterpriseIntegration</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=7453c5e92a8d" width="1" height="1" alt="">]]></content:encoded>
        </item>
    </channel>
</rss>