[Blog](https://supermemory.ai/blog) · [Engineering](https://supermemory.ai/blog/tag/engineering) # The UX and technicalities of awesome MCPs By Dhravya Shah June 8, 2025 · 7 min read ![Supermemory header reading "We made memory MCP easier to use" — One plug aligns with its matching port for a simple connection](https://supermemory.ai/_astro/cover.CynQFnvu_tnk3b.webp) Last month, we launched the Supermemory MCP, mostly to test our own infrastructure and get some initial traction. It blew up. To my absolute surprise, the initial launch itself got _**half a million impressions (!!!)**_. Then, we launched and got [#2 on ProductHunt](https://www.producthunt.com/products/supermemory?launch=universal-memory-mcp) too. Seems like people absolutely **loved** the concept and it was a massively successful side-project. In this blog, we'll talk about the reasons why this MCP did really well - and learnings along the way. X content is paused until you allow embedded content. [Open on X ↗](https://twitter.com/DhravyaShah/status/1912544775536066772?ref%5Fsrc=twsrc%5Etfw) > ## What did we do differently? If you look at any popular MCP server right now, the ultimate realizations are: * MCPs are only used by developers right now. * Authentication always seems to get in the way * It's a terrible pain to install MCPs today. * MCP clients always want to behave differently. * Long-standing connections (SSE) keeps breaking. * There's a huge usecase problem. To get through each of these hurdles, we chose a set of user experience and technical decisions, to ensure that the MCP is really _newbie proof_ (from both a user perspective and not-well-done MCP clients) ## 1) Choose SSE instead of STDIO At the time of building the MCP, most servers were using STDIO. But, the reality is, my mom could _never ever_ use something that needs a terminal command to run. I realized that there will be more and more MCP clients that would add SSE support, especially on Web-based clients like Claude.ai. In April there were barely any famous web-based clients (none that I can remember), so this was a huge bet. My previous MCP servers like [apple-mcp](https://github.com/supermemoryai/apple-mcp), which also got popularity, were completely local-only. This felt like a good bet to make because MCP SSE servers can be made to work over STDIO as well. ## 2) No auth at all. Next step was authentication: There's a lot of fuss in the "Auth with MCP" space, and in my opinion, none of them offer a good user experience. The main issue in my opinion is, again, the variability of how the clients work. So, we decided to nuke auth completely - and generate a unique URL for every user that comes to the website. In fact, we made it so easy that all that the user has to do is [mcp.supermemory.ai](https://mcp.supermemory.ai/), and the ONLY thing that they see on the screen is a huge URL to copy. From a technical perspective, we use react-router loader to: 1. Check if `userId` is in session 2. If it is, query the database for that userId 3. Else, generate a random ID 4. Commit it to the session ``` export async function loader({ request, context }: Route.LoaderArgs) { const cookies = request.headers.get("Cookie") const session = await getSession(cookies) if (session.has("userId")) { const userId = session.get("userId")! // ... get user memories return data({ message: "Welcome back!", userId, memories, }) } session.set("userId", nanoid()) const userId = session.get("userId")! return data( ... { headers: { "Set-Cookie": await commitSession(session, { expires: new Date("9999-12-31"), }), }, }, ) } ``` There are obviously some huge drawbacks to this approach: * If the user clears the cookies, they will lose their memories * If anyone finds out the URL for the user, they might be able to access it. These drawbacks were fine for us, because the user would "persist" their URL by Copy+pasting it to an MCP client anyways. Which means they would always store it _somewhere_. And the url path parameter acted as an API key. This also means that the MCP server itself would have to handle the userId on a _dynamic_ url. But we'll get there. ![Flowchart: if a userId is in cookies, query the Supermemory API for memories; if not, generate a random ID and commit the session](https://supermemory.ai/_astro/image-1.DRgn-Z1R_cfgdT.webp) How our authentication works This did make things a _little_ challenging - like how do you handle different client connections again on different domains? Most MCP server libraries are built for one URL only- so we "wrapped" our functions into a `createSupermemory` "meta mcp" so, technically, a dynamic server is generated for every single unique user. ``` server.get("/sse", async (c) => { const userId = c.get("userId") return streamSSE(c, async (stream) => { this.transport?.connectWithStream(stream) await bridge({ mcp: muppet(createSuperMemory(userId, c.env), { name: "Supermemory MCP", version: "1.0.0", }), transport: c.env.transport, // logger: logger, }) }) }) ``` This does not really have any performance implications, but just better UX because the URL will look like `mcp.supermemory.ai/userid/sse` and `/userid/messages` Beautiful, as things should be. ## 3) Made an installation CLI to make it dead simple. Over a quick afternoon, I vibecoded this quick tool to install MCPs with a single CLI command. This meant that I could provide a CLI command for whichever MCP client had a local-only config. Interestingly, a lot of cool people contributed to the CLI like [Namanya](https://nmn.gl/), [Kent C. Dodds](https://kentcdodds.com/), and more, to add more clients, and make the installation UX even better! ![Supermemory MCP install UI with client tabs (claude, cursor, cline, windsurf) and an npx install-mcp command](https://supermemory.ai/_astro/image-3.EXvUmnMe_Z2kVytD.webp) Users no longer had to "Edit JSON Config" files to install a simple MCP server. It was always Copy+paste. Either the URL to direct SSE connections, or the CLI command. Simply beautiful. As I like it. ![Terminal showing "Successfully installed MCP server" for supermemory-mcp with the claude client](https://supermemory.ai/_astro/image-4.BnpR6kZ1_1sMX6t.webp) This is what I mean by good UX. You need to go out of your way to give a good user experience. Software is flaky and broken, so the easier you make it for people to _not_ fuck up, the better. **Supermemory does all this because we care. Even for our non-revenue making projects :)** ## 4) Provide prompts. The supermemory MCP also provides a prompt built-in for clients that really, really don't want to listen. For example, when we launched, the MCP didn't work on Claude because it really wanted to use a resource instead of a tool. Which meant that we had to write an unhinged prompt 😆 ![addToSupermemory tool toggle with a description stressing it must be called as a tool, not a resource](https://supermemory.ai/_astro/image-5.CeVYY0u2_1UM3tO.webp) Anyways, by prompts, I mean actual MCP protocol "prompts". So that users can actually click on "use Supermemory Prompt" whenever they explicitly need to use memory ![Claude chat composer with a Supermemory Prompt.txt attachment and the Sonnet 4 model selected](https://supermemory.ai/_astro/image-6.DAAttVR__rSoLA.webp) This is not perfect, but all we could do to make MCP clients unfuck themselves :) ## 5) Host your MCP servers on Cloudflare. SSE is a weird protocol. You have to have an _**extremely**_ long-running connection for the messages to actually work. Memory absolutely requires this because clients may send one memory in 30 minutes, or 6 hours, you never know. Fortunately, the smart folks over at Cloudflare have been working on a way to run "Durable" objects / connections for years now. And MCP happens to be the perfect use case for something like this. Initially when we launched the MCP, users constantly complained about "Connection lost" and stuff like that. I was also lucky to work on this tech, during my time at Cloudflare before starting this company. So, I trust the [Agents SDK](https://agents.cloudflare.com/) and genuinely think that there's no better platform to put MCPs on. Worse still, most platforms charge for Execution time (total time running compute, including waits) - not CPU time (only time actually using CPU), but for MCPs you almost always have to just "wait" for messages to do anything with it. Again, fortunately Cloudflare only charges for CPU time which means I can have unimaginably long connections, and that _definitely_ happened. ![Two dashboards contrasting CPU time peaking near 5.67ms P99 against request duration spiking to ~1.54M ms P99](https://supermemory.ai/_astro/image-7.BWK5VOgL_A1Ikc.webp) While the actual CPU time was a few milliseconds only (mostly MCP transport stuff), the total connection times were MILLIONS of milliseconds each. Bonkers 😅 ## 6) Use-cases. On a more non-technical side, I believe that whenever you sell something, lead with examples. All the demos of the supermemory MCP were things that _I_ personally use it for, every day. In fact, my personal URL has 400+ memories about me already - projects I'm building, my preferences, my girlfriend's name, and a lot more. 😃 I made & posted videos of me actually using the product, and whenever something interesting happens, I would post it on Twitter as well! The result: growth. It turns out, more MCP clients started supporting direct SSE connections - include Claude, [x.com/DhravyaShah/status/19304649496772…](https://x.com/DhravyaShah/status/1930464949677240703) \- And instantly after Claude launched integrations, people started looking for interesting integrations to add. Supermemory was one of them, on day 1. Today, it might be the most used memory MCP - but I can't tell for sure 😅 But all this could not be possible without all the background work... X content is paused until you allow embedded content. [Open on X ↗](https://twitter.com/DhravyaShah/status/1931484491048169883?ref%5Fsrc=twsrc%5Etfw) > ## The background work We were able to make & ship supermemory MCP in \~5 hours of actual work time. Just look at the repository - it's the simplest thing you'll see in your life. It's a react router app that makes fetch calls to the Supermemory API. The MCP was made as a "Customer" of the base API, that we have been working for months on. If you want to read more about that, we wrote a blog about it too! But the TL;DR is that we've been obsessing over memory for a while now, topping almost every benchmark and objectively the best way to build chat apps / agents with memory today ![Bar chart: Supermemory beats the next best memory provider on P@1, Recall@10, NDCG@10, and MAP@10 in the LOCOMO benchmark](https://supermemory.ai/_astro/image-10.CFiVBJ0O_Z2emdr1.webp) Supermemory vs next best provider, in the LOCOMO benchmark. There's a lot to build, and we're just getting started. We'll continue obsessing over user and developer experience, and continue on our vision of building the best memory engine on the planet. Star the [MCP repository](https://github.com/supermemoryai/supermemory-mcp), start [using supermemory](https://docs.supermemory.ai/) and leave your thoughts on Twitter (tag [@supermemoryai](https://x.com/supermemoryai)) ## Other posts. 1. [We're open sourcing the company brain. Here's how we designed the multiplayer harness Company Brain is now open source. A walkthrough of the multiplayer harness behind its Slack experience, from proactivity and memory boundaries to approvals and recovery. NewsSep 25, 2026 ](https://supermemory.ai/blog/open-sourcing-company-brain) 2. [Jev changes a lot in memory & context engineering. Here's exactly how. We tested Jev across reranking, chunking, observation, and harness decisions. Here is where fast decision models help memory systems, where they cost more, and where they still fall short. EngineeringSep 24, 2026 ](https://supermemory.ai/blog/jev-memory-context-engineering) 3. [I reverse-engineered Instinct's memory. Here's exactly how it works Instinct keeps its memory as git-tracked markdown files, found with grep rather than vectors. Here is the whole system as far as black-box probing can reconstruct it, and how to rebuild it on supermemory in about 60 lines. EngineeringSep 20, 2026 ](https://supermemory.ai/blog/reverse-engineering-instinct-memory) 4. [An update to supermemory We've discontinued the supermemory company brain and Nova. Everyone who was charged has been refunded, our MCP and plugins continue to run, and we're going all in on the memory engine. NewsSep 10, 2026 ](https://supermemory.ai/blog/an-update-to-supermemory) 5. [Scaling Conversations: How Adapta Grew Usage Without Losing Context Adapta added Supermemory as a persistent memory layer so every conversation keeps its context — letting the team scale usage without losing the thread. Case StudyJun 12, 2026 ](https://supermemory.ai/blog/adapta-scaling-conversations) 6. [How Chatarmin Ditched RAG and Went Memory-Only with Supermemory Chatarmin replaced a heavy RAG pipeline with Supermemory's memory layer — cutting average AI response time from 40s to 12s and token usage by 40–50%. Case StudyJun 10, 2026 ](https://supermemory.ai/blog/chatarmin-memory-only) 7. [SMFS: making agentic retrieval 55% cheaper AND more accurate We launched SMFS.ai (Supermemory Filesystem) a few weeks ago, with a simple bet: We can redesign the filesystem specifically for agents, with special files, structures, and commands that it can use for it's tasks. Today, SMFS is used by hundreds of companies to power their agents. EngineeringMay 28, 2026 ](https://supermemory.ai/blog/smfs-making-agentic-retrieval-55-cheaper-and-more-accurate) 8. [Introducing Dynamic Dreaming: supermemory now connects the dots, for you. Dreaming is magical. TLDR: We're launching Dynamic Dreaming in supermemory today, which automatically works if you're using supermemory in any way - API, OpenClaw, Hermes agent, etc. EngineeringMay 25, 2026 ](https://supermemory.ai/blog/introducing-dynamic-dreaming-supermemory-now-connects-the-dots-for-you) 9. [Dear reader, we just made supermemory insanely cheap... the Context Cloud When I first started building supermemory, I had one goal: To build the best memory system for AI. I would talk to customers, and find out that memory was not the only thing they needed - They were all setting up 7-8 different vendors at the same time. EngineeringMay 18, 2026 ](https://supermemory.ai/blog/dear-reader-we-just-made-supermemory-insanely-cheap-the-context-cloud) 10. [Introducing @supermemory/tools v2.0.0 Today we're releasing v2.0.0\. This release unifies the API across all agents sdk integrations from AI SDK to Mastra, makes conversation identity a first-class concept, and ships with memory saving on by default. EngineeringApr 27, 2026 ](https://supermemory.ai/blog/introducing-supermemory-tools-v2-0-0) 11. [Solving the Precision-Recall Tradeoff: Search Result Aggregation When you're building memory for AI, search is your foundational layer. The way search generally works is straightforward: the user defines a query, and then sets a limit (top-K) on how many search results they want returned. Usually, this is set to 10 or 20. EngineeringApr 5, 2026 ](https://supermemory.ai/blog/solving-the-precision-recall-tradeoff-search-result-aggregation) 12. [OpenClaw Memory Problems: Why It Forgets and How to Fix It (2026) TLDR: Today, we are releasing a new version of our openclaw plugin - https://github.com/supermemoryai/openclaw-supermemory. This post is going to be a bit technical, so bear with me (or bookmark for later!) In this post, I will talk about what we do about OpenClaw memory, and how we fix it. EngineeringFeb 19, 2026 ](https://supermemory.ai/blog/why-everyone-is-complaining-about-openclaws-memory-it-sucks-and-why-supermemory-fixes-it) 13. [Stateful Coding Agents with Memory: Build Long-Running Agents (2026) We built a plugin for Claude Code and OpenCode that gives your coding agent persistent memory. It remembers your preferences, learns your codebase, and never loses context mid-conversation. The result is an agent you can run for months without starting over. EngineeringFeb 18, 2026 ](https://supermemory.ai/blog/infinitely-running-stateful-coding-agents) 14. [Clawd / Molt bot's memory SUCKS. We gave it supermemory. I'm the founder of supermemory. Clawd/Molt bot is blowing up right now, with many, many use cases. I set it up, too, and have been using it through telegram. TLDR: just go to https://supermemory.ai/docs/integrations/clawdbot to set up supermemory for your clawd bot. EngineeringJan 28, 2026 ](https://supermemory.ai/blog/clawd-molt-bots-memory-sucks-we-gave-it-supermemory) 15. [Catch up with our UNFORGETTABLE Launch Week Over the last year, one belief has guided almost everything we’ve built at Supermemory AI becomes meaningfully useful only when it remembers. Memory shouldn’t be something developers rebuild from scratch. It shouldn’t be fragile, expensive, or trapped inside a single tool. NewsJan 4, 2026 ](https://supermemory.ai/blog/catch-up-with-our-unforgettable-launch-week) 16. [Empowering the Next Generation of Founders: Supermemory Startup Program If there’s one thing we’ve learned while building Supermemory, it’s that most startups don’t fail because they didn't build features; they fail when infrastructure slows them down, or they built too slow. NewsDec 31, 2025 ](https://supermemory.ai/blog/empowering-the-next-generation-of-founders-supermemory-startup-program) 17. [Building code-chunk: AST Aware Code Chunking At Supermemory, we're building context engineering infrastructure for AI. A huge part of that is dealing with code: ingesting repos, understanding structure, and making it searchable. The problem is that most code chunking solutions are terrible. We built code-chunk to fix this. EngineeringDec 29, 2025 ](https://supermemory.ai/blog/building-code-chunk-ast-aware-code-chunking) 18. [Supermemory raises $3 million with the best memory engine for LLMs Today, I am excited to announce our first funding round to accelerate our mission of building an interoperable, scalable and reliable memory for LLMs and agents. Memory is one of the hardest challenges in AI right now. NewsOct 6, 2025 ](https://supermemory.ai/blog/supermemory-raises-3-million-and-building-the-best-memory-engine-for-llms) 19. [Mem0 vs Supermemory: Why Scira Switched Scira AI moved its production memory layer from Mem0 to Supermemory. This is what failed, what improved, and how the team evaluated the two systems. Case StudyOct 2, 2025 ](https://supermemory.ai/blog/why-scira-ai-switched) 20. [Never Record Again: How Montra Uses Supermemory to Rethink Video Creation Campbell Baron, the founder of Montra, has been making videos since he was twelve. By thirteen, he was already doing brand work. Today, he’s betting on a very different future for creators: a world where recording is the exception, and most videos are generated from scratch. Case StudyAug 21, 2025 ](https://supermemory.ai/blog/never-record-again-how-montra-uses-supermemory-to-rethink-video-creation) 21. [Unified Memory That Works Where You Work: Your Second Brain With Supermemory Hi everyone, I’m Dhravya, the founder of Supermemory. I want to start with a little story behind why this product means so much to me. You can also skip straight to what it is and how it works below. EngineeringJul 25, 2025 ](https://supermemory.ai/blog/unified-memory-that-works-where-you-work-your-second-brain-with-supermemory) 22. [Supermemory just got faster on PlanetScale What is Supermemory? Supermemory completes the missing part of the LLM puzzle: memory. Just as memory is crucial for human intelligence, it's essential for truly intelligent AI systems. EngineeringJul 18, 2025 ](https://supermemory.ai/blog/supermemory-just-got-faster-on-planetscale) 23. [Faster, smarter, reliable infinite chat: Supermemory IS context engineering. People are obsessed with prompts and prompt engineering. Sure, what you say is important, but what the model knows when you say it is the difference between a stateless text generator and an intelligent AI system. In short, context is the most crucial component. NewsJul 9, 2025 ](https://supermemory.ai/blog/faster-smarter-reliable-infinite-chat-supermemory-is-context-engineering) 24. [We solved AI API interoperability One API to rule them all, One spec to find them, One library to bring them all and in the TypeScript, bind them. When we were building the Infinite Chat API, initially, we only supported the OpenAI format. This was fine, until a lot of our customers started asking for more. EngineeringJul 7, 2025 ](https://supermemory.ai/blog/we-solved-ai-api-interoperability) 25. [The Wow Factor of Memory - How Flow Used Supermemory To Build Smarter, Stickier Products Overview: Flow is a note-taking app built around a bold vision: to create a more personal, context-aware writing experience powered by AI. At the heart of this mission is memory. Case StudyJun 14, 2025 ](https://supermemory.ai/blog/the-wow-factor-of-memory-how-flow-used-supermemory-to-build-smarter-stickier-products) 26. [Architecting a memory engine inspired by the human brain Language is at the heart of intelligence, but what truly powers meaningful interaction is memory — the ability to accumulate, recall, and contextualize information over time. Large Language Models (LLMs) have mastered language, but memory remains their Achilles’ heel. EngineeringJun 5, 2025 ](https://supermemory.ai/blog/memory-engine) ## Start building with supermemory. Memory and continual learning for any model, any harness. Available through our API, plugins, and MCP. [Build with supermemory ](https://console.supermemory.ai/)