Piecing together the Agent puzzle: MCP, authentication & authorization, and Durable Objects free tier

It’s not a secret that at Cloudflare we are bullish on the future of agents. We’re excited about a future where AI can not only co-pilot alongside us, but where we can actually start to delegate entire tasks to AI. 

While it hasn’t been too long since we first announced our Agents SDK to make it easier for developers to build agents, building towards an agentic future requires continuous delivery towards this goal. Today, we’re making several announcements to help accelerate agentic development, including:

  • New Agents SDK capabilities: Build remote MCP clients, with transport and authentication built-in, to allow AI agents to connect to external services. 

  • BYO Auth provider for MCP: Integrations with Stytch, Auth0, and WorkOS to add authentication and authorization to your remote MCP server. 

  • Hibernation for McpAgent: Automatically sleep stateful, remote MCP servers when inactive and wake them when needed. This allows you to maintain connections for long-running sessions while ensuring you’re not paying for idle time. 

  • Durable Objects free tier: We view Durable Objects as a key component for building agents, and if you’re using our Agents SDK, you need access to it. Until today, Durable Objects Continue reading

Cloudflare acquires Outerbase to expand database and agent developer experience capabilities

I’m thrilled to share that Cloudflare has acquired Outerbase. This is such an amazing opportunity for us, and I want to explain how we got here, what we’ve built so far, and why we are so excited about becoming part of the Cloudflare team.

Databases are key to building almost any production application: you need to persist state for your users (or agents), be able to query it from a number of different clients, and you want it to be fast. But databases aren’t always easy to use: designing a good schema, writing performant queries, creating indexes, and optimizing your access patterns tends to require a lot of experience. Add that to exposing your data through easy-to-grok APIs that make the ‘right’ way to do things obvious, a great developer experience (from dashboard to CLI), and well… there’s a lot of work involved.

The Outerbase team is already getting to work on some big changes to how databases (and your data) are viewed, edited, and visualized from within Workers, and we’re excited to give you a few sneak peeks into what we’ll be landing as we get to work.

Database DX

When we first started Outerbase, we saw how Continue reading

Cloudflare Workflows is now GA: production-ready durable execution

Betas are useful for feedback and iteration, but at the end of the day, not everyone is willing to be a guinea pig or can tolerate the occasional sharp edge that comes along with beta software. Sometimes you need that big, shiny “Generally Available” label (or blog post), and now it’s Workflows’ turn.

Workflows, our serverless durable execution engine that allows you to build long-running, multi-step applications (some call them “step functions”) on Workers, is now GA.

In short, that means it’s production ready —  but it also doesn’t mean Workflows is going to ossify. We’re continuing to scale Workflows (including more concurrent instances), bring new capabilities (like the new waitForEvent API), and make it easier to build AI agents with our Agents SDK and Workflows.

If you prefer code to prose, you can quickly install the Workflows starter project and start exploring the code and the API with a single command:

npm create cloudflare@latest workflows-starter -- 
--template="cloudflare/workflows-starter"

How does Workflows work? What can I build with it? How do I think about building AI agents with Workflows and the Agents SDK? Well, read on.

Building with Workflows

Workflows is a durable execution engine built on Cloudflare Workers that Continue reading

Introducing AutoRAG: fully managed Retrieval-Augmented Generation on Cloudflare

Today we’re excited to announce AutoRAG in open beta, a fully managed Retrieval-Augmented Generation (RAG) pipeline powered by Cloudflare, designed to simplify how developers integrate context-aware AI into their applications. RAG is a method that improves the accuracy of AI responses by retrieving information from your own data, and providing it to the large language model (LLM) to generate more grounded responses.

Building a RAG pipeline is a patchwork of moving parts. You have to stitch together multiple tools and services — your data storage, a vector database, an embedding model, LLMs, and custom indexing, retrieval, and generation logic — all just to get started. Maintaining it is even harder. As your data changes, you have to manually reindex and regenerate embeddings to keep the system relevant and performant. What should be a simple “ask a question, get a smart answer” experience becomes a brittle pipeline of glue code, fragile integrations, and constant upkeep.

AutoRAG removes that complexity. With just a few clicks, it delivers a fully-managed RAG pipeline end-to-end: from ingesting your data and automatically chunking and embedding it, to storing vectors in Cloudflare’s Vectorize database, performing semantic retrieval, and generating high-quality responses using Workers AI. AutoRAG continuously monitors Continue reading

Welcome to Developer Week 2025

We’re kicking off Cloudflare’s 2025 Developer Week — our innovation week dedicated to announcements for developers.

It’s an exciting time to be a developer. In fact, as a developer, the past two years might have felt a bit like every week is Developer Week. Starting with the release of ChatGPT, it has felt like each day has brought a new, disruptive announcement, whether it’s new models, hardware, agents, or other tools. From late 2024 and in just the first few months of 2025, we’ve seen the DeepSeek model challenge assumptions about what it takes to train a new state-of-the-art model, MCP introduce a new standard for how LLMs interface with the world, and OpenAI’s o4 model Ghiblify the world.

And while it’s exciting to witness a technological revolution unfold in front of your eyes, it’s even more exciting to partake in it. 

A new era of innovation

One of the marvels of the recent AI revolution is the extent to which the cost of experimentation has gone down. Ideas that would have taken whole weekends, weeks, or months to build can now be turned into working code in a day. You can vibe-code your way through things you might Continue reading

AI for Network Engneers: Challenges in AI Fabric Design

 Introduction


Figure 10-1 illustrates a simple distributed GPU cluster consisting of three GPU hosts. Each host has two GPUs and a Network Interface Card (NIC) with two interfaces. Intra-host GPU communication uses high-speed NVLink interfaces, while inter-host communication takes place via NICs over slower PCIe buses.

GPU-0 on each host is connected to Rail Switch A through interface E1. GPU-1 uses interface E2 and connects to Rail Switch B. In this setup, inter-host communication between GPUs connected to the same rail passes through a single switch. However, communication between GPUs on different rails goes over three hops  Rail–Spine–Rail switches.

In Figure 10-1, we use a data parallelization strategy where a training dataset is split into six micro-batches, which are distributed across the GPUs. All GPUs use the shared feedforward neural network model and compute local model outputs. Next, each GPU calculates the model error and begins the backward pass to compute neuron-based gradients. These gradients indicate how much, and in which direction, the weight parameters should be adjusted to improve the training result (see Chapter 2 for details).

Figure 10-1: Rail-Optimized Topology.


Egress Interface Congestions


After computing all gradients, each GPU stores the results in a local memory buffer and Continue reading

Meta’s Llama 4 is now available on Workers AI

As one of Meta’s launch partners, we are excited to make Meta’s latest and most powerful model, Llama 4, available on the Cloudflare Workers AI platform starting today. Check out the Workers AI Developer Docs to begin using Llama 4 now.

What’s new in Llama 4?

Llama 4 is an industry-leading release that pushes forward the frontiers of open-source generative Artificial Intelligence (AI) models. Llama 4 relies on a novel design that combines a Mixture of Experts architecture with an early-fusion backbone that allows it to be natively multimodal.

The Llama 4 “herd” is made up of two models: Llama 4 Scout (109B total parameters, 17B active parameters) with 16 experts, and Llama 4 Maverick (400B total parameters, 17B active parameters) with 128 experts. The Llama Scout model is available on Workers AI today.

Llama 4 Scout has a context window of up to 10 million (10,000,000) tokens, which makes it one of the first open-source models to support a window of that size. A larger context window makes it possible to hold longer conversations, deliver more personalized responses, and support better Retrieval Augmented Generation (RAG). For example, users can take advantage of that increase to summarize multiple documents or Continue reading

io_uring, kTLS and Rust for zero syscall HTTPS server

Around the turn of the century we started to get a bigger need for high capacity web servers. For example there was the C10k problem paper.

At the time, the kinds of things done to reduce work done per request was pre-forking the web server. This means a request could be handled without an expensive process creation.

Because yes, creating a new process for every request was something perfectly normal.

Things did get better. People learned how to create threads, making things more light weight. Then they switched to using poll()/select(), in order to not just spare the process/thread creation, but the whole context switch.

I remember a comment on Kuro5hin from anakata, the creator of both The Pirate Bay and the webserver that powered it, along the lines of “I am select() of borg, resistance is futile”, mocking someone for not understanding how to write a scalable webserver.

But select()/poll() also doesn’t scale. If you have ten thousand connections, that’s an array of ten thousand integers that need to be sent to the kernel for every single iteration of your request handling loop.

Enter epoll (kqueue on other operating systems, but I’m focusing Continue reading

From Python to Go 018. Interaction With Network Devices Using GNMI.

Hello my friend,

Within past blog posts we covered how to interact with network devices (in fact, many servers support that as well) using SSH/CLI and NETCONF/YANG. Those two protocols allow you to confidently cover almost all cases for managing devices, whether you prefer more human-like approach, that is templating CLI commands and pushing them via SSH or using structured XML data and send it via NETCONF. However, there are more protocols and today we are going to talk about the most modern to the date, which is called GNMI (generalized network management interface).

How Does Automation Help Real Business?

Talking to students at our trainings and with customers and peers at various events, there is often a concern arise that small and medium businesses don’t have time and/or need to invest in automation. There is no time as engineers are busy solving “real” problems, like outages, customer experience degradation, etc. There is no need because why to bother, we have engineers to tackle issue. From my experience automation allows to save so much time on doing operational tasks and to free it up for improving user experience and developing new problems, that is difficult to overestimate. It really is Continue reading

Querying Data in Infrahub via the Python SDK

Querying Data in Infrahub via the Python SDK

Infrahub provides multiple ways to interact with your infrastructure data, including the Web GUI, GraphQL queries, and the Python SDK. These can be used to query, modify, create, or delete data in Infrahub. In this post, we’ll focus on using the Python SDK to query data from Infrahub.

This post assumes you are familiar with basic Python and Infrahub. If you’re new to these topics, don’t worry, you can still follow along.

Throughout this post, we’ll be using the Always-On Infrahub demo instance, which is available for anyone to access via this link. The demo instance already has some data in it, so if you’d like to follow along or try this yourself, you can use it without needing to set up anything.

Originally published under - https://www.opsmill.com/querying-data-in-infrahub-via-the-python-sdk/

Introduction

The Python SDK supports both synchronous and asynchronous Python. However, in this post, we’ll focus on using synchronous Python, which I hope most of us are comfortable with. We’ll cover async in a future blog post.

Interacting with Infrahub through the Python SDK is done using a client object, which defines the Infrahub instance you’ll be working with. This client acts as the connection point, allowing you to Continue reading

Dave Taht, Who Sped Up Networks More Than You’ll Ever Know, Has Died

Dave Taht, LinkedIn profile photo. I don’t recall when I first met InterOp conference in the 80s when it was the best-ever networking conference or at a science-fiction convention — Balticon? — around the same time. Whether it was when he was talking about TCP/IP networking or playing Bufferbloat Project with Jim Gettys, focusing on reducing Internet latency and improving network performance. His work on advanced queuing algorithms like Common Applications Kept Enhanced (CAKE) significantly enhanced network efficiency, making these technologies part of the default networking stack in many Linux distributions, the popular open source embedded router operating system,

Hedge 265: Out of Band Networks

Out of band management networks were once more common than they are today. Should we go back to building out of band management networks? Should out of band management networks be virtual or physical? How can we sell out of band management networks to the folks paying the bills? Daryll Swer joins Tom Ammon and Russ White to discuss the importance of OOB management.

download

TNO023: Networking’s Third Phase – The Network Operator Experience

Guest Chris Grundemann believes that NetOps is in the third phase of networking–improving the network operator experience. Not just making the network functional or improving end user experience. In this episode, Chris tells his origin story at a wireless service provider and growth into a founder of multiple companies. He also shares his community-focused work... Read more »

Cloudflare’s commitment to CISA Secure-By-Design pledge: delivering new kernels, faster

As cyber threats continue to exploit systemic vulnerabilities in widely used technologies, the United States Cybersecurity and Infrastructure Agency (CISA) produced best practices for the technology industry with their Secure-by-Design pledge. Cloudflare proudly signed this pledge on May 8, 2024, reinforcing our commitment to creating resilient systems where security is not just a feature, but a foundational principle.

We’re excited to share and provide transparency into how our security patching process meets one of CISA’s goals in the pledge: Demonstrating actions taken to increase installation of security patches for our customers.

Balancing security patching and customer experience 

Managing and deploying Linux kernel updates is one of Cloudflare’s most challenging security processes. In 2024, over 1000 CVEs were logged against the Linux kernel and patched. To keep our systems secure, it is vital to perform critical patch deployment across systems while maintaining the user experience. 

A common technical support phrase is “Have you tried turning it off and then on again?”.  One may be  surprised how often this tactic is used — it is also an essential part of how Cloudflare operates at scale when it comes to applying our most critical patches. Frequently restarting systems exercises the Continue reading

Transitioning into Networking, 2025 Edition

Elmer sent me the following question:

I’ve been working in systems engineering (Linux, virtualization, infrastructure ops) and am considering shifting toward network engineering or architecture. I got my CCNA years ago and started CCNP but didn’t continue.

I’d really appreciate any thoughts you might have on how someone with my background could best make that transition today, especially with how things are evolving around automation and the cloud.

I keep answering a variant of this question every other year or so (2019, 2021, 2023, 2024). I guess it’s time for another answer, so here we go.

Read more …

KubeCon Europe: Kgateway Aims To Be the Kubernetes Onramp

Kubernetes network administrators at KubeCon + CloudNativeCon EU this week in London should drop by the ease the management of moving traffic to and from clusters. Built on top of Kubernetes Gateway API, the open source Solo.io, and went under the name Gloo Gateway. At last year’s KubeCon +_ CloudNativeCon North America 2024, the company announced that it would be donating the software to the Cloud Native Computing Foundation (CNCF), changing the software’s name to kgateway in the process. In March, CNCFGloo open source repository will be deprecated over time. The Importance of the Kubernetes Gateway API In 2023, the

1 … 70 71 72 73 74 … 3,849