TE
科技回声
首页24小时热榜最新最佳问答展示工作
GitHubTwitter
首页

科技回声

基于 Next.js 构建的科技新闻平台,提供全球科技新闻和讨论内容。

GitHubTwitter

首页

首页最新最佳问答展示工作

资源链接

HackerNews API原版 HackerNewsNext.js

© 2025 科技回声. 版权所有。

Show HN: Klavis AI – Open-source MCP integration for AI applications

79 点作者 wirehack3 天前
Hi HN, we are excited to show you Klavis AI. It is an open source project and we provide hosted versions with API access as well. (Website: <a href="https:&#x2F;&#x2F;www.klavis.ai&#x2F;">https:&#x2F;&#x2F;www.klavis.ai&#x2F;</a>, Github repo: <a href="https:&#x2F;&#x2F;github.com&#x2F;Klavis-AI&#x2F;klavis">https:&#x2F;&#x2F;github.com&#x2F;Klavis-AI&#x2F;klavis</a>)<p>We&#x27;re addressing a couple of key problems with using MCPs. First, many available MCP servers lack native or used-based authentications, creating security vulnerabilities and adding complexity during development.<p>Second, many MCP servers are personal projects, not designed for the reliability needed in production.<p>Connecting to these servers usually requires writing custom MCP client code for the MCP protocol itself, which is a barrier, especially if you already have function calling systems in place.<p>Klavis AI aims to address these issues. To simplify access, we offer an API to launch production-ready, hosted MCP servers quickly via our API. The API also provides built-in OAuth and multi-tenancy auth support for MCP servers.<p>We also want to remove the need for developers to write MCP client code. You can use our API to interact with any remote MCP servers directly from your existing backend infrastructure. For faster prototyping or direct user interaction, we also provide open-source client interfaces for Web, Slack, and Discord.<p>The MCP servers and clients code is open source because we want to contribute to the MCP community.<p>For a quick start in the hosted verions, log in to our website and generate an API key. Then start calling our APIs directly. You can find more details in our doc: <a href="https:&#x2F;&#x2F;docs.klavis.ai">https:&#x2F;&#x2F;docs.klavis.ai</a><p>For a quick start in the open source version, go to our github repository and check out the detailed readme on each MCP server and client.<p>A little note about myself: my background includes working on the function calling for Google Gemini. During that time, I saw firsthand the challenges teams face when trying to connect AI agents to external tools. I want to bring my insights and energy to accelerate MCP adoption.<p>This is an early release, and we’d appreciate feedback from the community. What are your worst pain points related to MCPs, either as a developer or a general user? What other MCP servers or features would be most valuable to you?<p>We&#x27;ll be around in the comments. Thanks for reading!

13 条评论

barefeg3 天前
For some time, authentication was not part of the MCP. Now it’s there <a href="https:&#x2F;&#x2F;modelcontextprotocol.io&#x2F;specification&#x2F;2025-03-26&#x2F;basic&#x2F;authorization" rel="nofollow">https:&#x2F;&#x2F;modelcontextprotocol.io&#x2F;specification&#x2F;2025-03-26&#x2F;bas...</a> so I’m wondering what is being addressed in Klavis. Is it something that the reference implementation of MCP lacks? If so, will it eventually make it to MCP?<p>I think it’s important to release SDKs that are secure by default, so not providing this in the reference MCP would be a big issue.<p>In my view, MCP should be maintained by the vendors themselves. It’s too complicated to use in the enterprise if everything comes from the community with questionable security. So I applaud initiatives that try to solve this. I think smithery.ai provides something similar while also being a repository of servers (I’m not associated with them), but again the problem is needing to trust an extra middleman vendor.<p>Does anyone else share this view? For example, will AWS (or insert any other hyperscaler) end up providing the “Bedrock” of MCP where security is native to the platform? Or will individual companies (Box, Google, MS, etc.) start rolling them out as part of their standard developer APIs?
评论 #43897465 未加载
评论 #43899137 未加载
BrandiATMuhkuh3 天前
Looking great!<p>I&#x27;m actually actively looking for something like this. But I&#x27;m not sure it fulfills my requirements.<p>Here is what I&#x27;m looking for - curated open source MCP servers - trusted curator - direct oauth integration, so the user permission is taken over (e.g.: you login with your companies Microsoft account and not with the account of the AI-Client.) this is a must for enterprise - easy to ship (eg.: via electron app) - I as the developer or the company owner can decide what servers are available to the user, and what tools (e.g.: most companies I talk to want read only access for their employees)<p>If Klavia fulfills that I would like to have a chat :)
评论 #43897648 未加载
pzo3 天前
This looks great will definitely have a look. What I wish that there was also MCP client (SDK) for mobile devices either native (Swift &#x2F; Kotlin ) or React Native &#x2F; Expo ( ideally with integration with vercel ai sdk). For mobile dev this would simplify setup (not need to have a proxy server) and allow business models with end-user Bring-Your-Own-Key.
评论 #43898070 未加载
TZubiri3 天前
This looks like it would add nothing. MCP&#x27;s were released like 2 seconds ago, if you want to build something in prod, you can build it against the official MCP code.<p>If you want to add something in the middle or a MCP specific dependency it should add some value, some specific value: industry specific, or add some specific feature.<p>This is just a generic dependency that provides overly generic features like &quot;production ready&quot; and &quot;integration&quot; (that&#x27;s true for all dependencies.<p>Par for the course for AI companies, &quot;do everything for everyone&quot;
评论 #43902780 未加载
danenania3 天前
This looks interesting! I&#x27;ll be keeping an eye on it. A few questions that come to mind:<p>- Is there a way to run it locally&#x2F;self-host?<p>- Are there discovery endpoints to list the available servers?<p>- The &#x27;Test &amp; Eval&#x27; page is interesting to me, as I think unpredictability of results with multiple MCP prompts&#x2F;sources interacting is generally a pretty big issue with MCP, so a good integrated eval system could be valuable. I see it&#x27;s not launched yet, but it would be great to know more about how this will work.
评论 #43897235 未加载
评论 #43896724 未加载
cloudking3 天前
What is the UX for an end user with the OAuth? They use MCP say for GitHub in Cursor, how do they provide access to their GitHub?<p>Also why would they not just use the official GitHub MCP?
评论 #43897549 未加载
评论 #43897856 未加载
orliesaurus3 天前
If it&#x27;s open source, why do I need an API key to get started? What does the API key do?
评论 #43897104 未加载
评论 #43897082 未加载
atonse3 天前
Does this mean that we&#x27;d have to store our various keys&#x2F;credentials with you (in the hosted solution).<p>If so, I would emphasize or add more documentation on how you are securing credentials.<p>And how would I add this to my agent? Would you have One MCP To Rule Them All?
评论 #43897505 未加载
srameshc3 天前
I like from what I see and the option of self hosting. Does it have a middleware like feature to extend, say I want to add more apps than what is provided out of the box ?
评论 #43897446 未加载
gitroom3 天前
Honestly, feels like every time I look at this stuff, the basics like consistent auth and real standards are still kinda up in the air. I keep thinking the endless remixing happens because nobodys happy with defaults yet. You think well ever land on one real standard or is everyone just gonna keep rolling their own forever?
评论 #43899796 未加载
Nedomas3 天前
Welcome to MCP hosting space guys. We’ve also been hosting MCPs since early Feb in <a href="https:&#x2F;&#x2F;supermachine.ai" rel="nofollow">https:&#x2F;&#x2F;supermachine.ai</a> and investing a lot into open-source (see <a href="https:&#x2F;&#x2F;github.com&#x2F;supercorp-ai&#x2F;supergateway">https:&#x2F;&#x2F;github.com&#x2F;supercorp-ai&#x2F;supergateway</a> and many others).<p>Guess there’s gonna be more competition haha (but don’t worry, I think our approaches are a bit different).
评论 #43899609 未加载
评论 #43902770 未加载
rc_mob3 天前
What is MCP?
评论 #43897395 未加载
评论 #43896871 未加载
joshstrange3 天前
I&#x27;ve seen a number of projects re-implement the same MCP servers that exist elsewhere, and I wonder if we&#x27;re doomed to do this over and over and over again. One of the huge draws of MCP servers was a reusable standard to describe tools, but if everyone re-implements their own tools, then where&#x27;s the value?<p>I assume this comes from the fact that it is very easy to create an MCP server, it is much harder to create a good MCP server. And so, a number of companies have fallen into &quot;not invented here&quot; or they just want to be able to control the quality of the MCP servers (that&#x27;s fair). a &quot;GitHub MCP&quot; is a dime-a-dozen but I wrote a lot of MCP servers just for myself because often the servers provided 80% of what I wanted. It&#x27;s early days for MCP but it&#x27;s been easier in all cases to just write my own MCP from scratch instead of trying to fork off something existing. I&#x27;m sure that will change as time goes on but MCP is, often, a thin layer over a SDK&#x2F;API, it&#x27;s not hard to implement.<p>All of this gets to an idea that I&#x27;ve been playing around in my head and and have written about a fourth of a blog post on, which is &quot;MCP interfaces&quot;. MCP Interfaces, put very simply, is the concept of a pluggable way to swap out different MCP servers that may not implement the exact same tools, but are the same spirit of tools. So, for example, a way to plug in Google search, Bing search, SearXNG, or Jina.ai search.<p>Why do we even need &quot;interfaces&quot;? Well, throwing all the tools at every LLM is a recipe for disaster in multiple ways (privacy, security, sanity, the list goes on) and often agents work best with a limited set of tools they know how to best utilize. That&#x27;s all well and good and I&#x27;m sure most SaaS&#x27; out there will write their own tools or pick off-the-shelf MCP servers but for the open source world I&#x27;ve been thinking it would be nice to slot in your own MCP tools for certain tasks.<p>Search is an easy example. Lots of tools can make use of search (other good examples might be headless&#x2F;headed browser automation, or memory&#x2F;RAG) and I&#x27;ve seen the trend of different LLM application asking for all sorts of ways to do search. Some ask for your Google API key, some want Jina, some want SearXNG (some want to spin up their own docker container for SearXNG). It&#x27;s all over the place, it&#x27;s inconsistent, and it&#x27;s a ton of wasted work. I don&#x27;t want to ask the developer of &quot;insert cool LLM application&quot; to support my special-snowflake MCP server or tool, but I also don&#x27;t want to just throw additional tools at the agents.<p>So far the best I can come up with is &quot;MCP Interface&quot; where an application can have a &quot;minimum viable&quot; tool(s) definition and then you can slot in MCP tools that meet that standard. Maybe &quot;Interface&quot; is too strict even, maybe it would just work to say &quot;Here are my search tools, use them wherever you would provide an agent with search features&quot;. &quot;MCP Tagging&quot;?<p>I&#x27;m not sure and in a year maybe we won&#x27;t even be talking about MCP or maybe the world of plug and play will be tiny&#x2F;non-existent so this idea will not matter. In a way it seems too utopian to imagine we&#x27;d end up in a place where a user can say &quot;I have my own search, let&#x27;s use that&quot; and instead we will have MCP app stores inside SaaS products (with no overlap or use outside the SaaS in question), or maybe we won&#x27;t even get that, you&#x27;ll just get whatever the SaaS wrote or, more likely, got from the open source community.
评论 #43899833 未加载