DevRev

You can join our Discord community here - discord.gg/d84dyzACef

Inside the Agent SDK: How we debugged the Claude Agent SDK x.com/i/broadcasts/1…

Work softer with Remote MCP @devrev 🌻 DevRev's MCP tools are design to be agnostic of your agent harness. This week Gokul shared with us some of the amazing tools that will help us to create issues, update comments and manage sprints! #mcp #devrev

So here's what happened. We were building DevRev's Agent SDK on top of the Claude Agent SDK. Things were going well – until they weren't. The TypeScript implementation didn't give us enough visibility into what was actually happening at runtime. Debugging felt like guessing. Integrations kept breaking in ways that were hard to explain. So we did what any reasonable engineering team would do – we started intercepting our own traffic. Reverse proxies. MITM-style tracing. Request interception. Basically, we treated the SDK like a black box and built the tools to see inside it. And honestly? The lessons we learned don't just apply to the Claude Agent SDK. They apply to anyone building on top of any agentic framework where the docs only tell you half the story. We're sharing the whole thing – live, unfiltered – on this DevRev Live. 📅 May 14th · 8:30 PM IST · Live on YouTube: ow.ly/ilwb50YXW8i

Monday morning. Inbox full. Slack buzzing. 14 "urgent" messages. Was it always like this? What if it's not really urgent at all? What if it's just… a lack of context. Your team not sharing the same information. Not knowing what other people know. That's why Mondays feel so loud. Computer, by DevRev, quietens all this noise. As the only AI with native "shared memory", Computer remembers everything. All your chats & decisions. All your docs & deals. Everything. Even better: it gets smarter every day. Other AIs forget – and guess. Computer remembers – and knows. Start your week softer: devrev.ai/work-softer

Ready for AI: design system for the agentic era x.com/i/broadcasts/1…

CACES Singapore 2026 is around the corner – and we'll be there as a co-presenting sponsor at Booth #4. Every day, brilliant teams are spending their energy on work that doesn't need to be this hard. More apps, more tools, more dashboards – but somehow, things still feel disconnected. We think it doesn't have to be that way. We built Computer to be the kind of teammate every team deserves – one that listens, learns, and actually helps get things done. We'd love for you to come see it. Swing by the booth, sit in on a session, or just come say hi. Some of the best conversations happen over coffee anyway. Know more: ow.ly/J3F650YVTkq

We just earned our ISO/IEC 27001:2022 certification – the global standard for information security management. But while the loudest debates in enterprise AI are about speed and capabilities – how fast can it reason? how much can it act? – the more important question has always been: “Can we trust this AI? Like, really, really trust it?” Computer earns its value by deeply understanding your business data – and a lot of that data is sensitive. ISO 27001 is the external validation that DevRev has the security discipline to handle that responsibly. From risk assessment to vendor management, incident response to employee security training – we are fully audited, and fully verified. Enterprise-grade security built in from the start, not bolted on as an afterthought. And that's why we're thrilled. If you're choosing an AI to work alongside your most sensitive workflows, this is the kind of foundation you need. The safety that lets you move fast – without looking over your shoulder. 🔗 Read more: shorturl.at/EUvcm

Your design system probably started with good intentions. Then someone skipped a component, engineering went rogue, and your brand deck slowly became a crime scene. We've been there. So we fixed it. Join us to know more. 🗓 May 7 · 8:30 PM IST 📺 Set a reminder and join us live: shorturl.at/jviIN #DevRev #DevRevLive

Every enterprise AI interaction in 2026: User: "What's happening with this account?" AI: checks Salesforce only "Everything looks great." Jira: has a P0 blocker filed yesterday Slack: champion vented at 11pm Zendesk: 3 open tickets The AI isn't lying. It just doesn't know what it doesn't know 🙃

Completely agree. And the bottleneck isn't engineering talent or model capability but it's that most products are still treating AI as a feature bolted onto existing workflows rather than rethinking what the workflow should be. The models can reason, take actions, hold context across sessions. Most products still use them to summarize text and autocomplete fields. The gap isn't 10 years of building. It's 10 years of unlearning how we designed software before this.

Recently heard someone say that if all LLM development/research suddenly froze, it would still take ~10 years for products to catch up to what the technology can do & this sounds about right to me

This is exactly right. The "one agent per workflow" approach is just the old integration spaghetti with a newer label. The part most people miss -- the spine isn't just an orchestration layer. It needs memory. If your AI doesn't already know how your customers, deals, tickets, and teams connect across systems before anyone asks a question and then every agent is still fetching context from scratch every time it runs. That's where the cost and the inaccuracy live. The spine has to understand the business, not just route data between tools.

If you are introducing your company or department to AI, do not add to your software bloat. This is the number one failure mode that I see plaguing companies today. Your finance department, for example, has 100+ workflows, and if you have a separate agent/automation for every single one of those, you're creating a tech-debt hell-hole that is impossible to dig yourself out of. Instead, approach AI agents from the key principle of on-top and in-between. That means 1. have a single pane of glass over all of your existing software, where the AI bubbles insights to the top, and 2. have your AI agents that run each individual software piece, passing data back and forth between them with high accuracy. You should have an AI “spine” that all agents live on top of, and this is the #1 reason why vibe coding tools like Lovable and Replit will never bring background agent ROI to enterprise.

5 minutes… that turns into 15… then 30… Then that “quick check-in?!” has ruined your afternoon. It’s not that people aren’t working hard enough. It’s that the work around the work keeps growing. Check-ins to prep for check-ins. Status updates about the status meeting, which needs a status report, which you haven’t had time to work on. Because of all the meetings. All because no one can trust the AI answers they’re getting. You need to check. Re-check. Triple-check. Computer, by DevRev, is the only AI with native “shared memory” built in. Which means you, and your whole team, gets AI answers they can trust. Every time. Less back-and-forth. More getting on with it. It’s time to work softer: devrev.ai/work-softer #WorkSofter

Most rooms in the Bay are built around speakers. Ours is built around builders. For AIBoomi Bootcamp ’26 | Bay Area, we’re bringing together: ⭐ @aparnadhinak | @arizeai ⭐ @HasijaKanav | @MeltPlanAI ⭐ @DevReveler | @devrev ⭐ @nhlhomer | @OpenAI ⭐ @prashantmital | @OpenAI ⭐ @prukalpa | @AtlanHQ ⭐ @rajoshighosh | @PromptQL ⭐ @srikrishnang | @RocketlaneHQ ⭐ @srinivasnjay | @interfaceAI ⭐ @swapnil | @observeAI ⭐ @amnigos | @atomicworkhq And many more. Founders. Operators. People who’ve actually built. Fused with real conversations and breakdowns. Including one full day inside OpenAI. 📍 May 18–20 · San Francisco If this is your kind of room, you should be in it >> aiboomi.org/events/ai-boot… #MayInTheBay #AI #AIBoomi

AI is a human problem, not a technology problem. The companies that win will be the ones that design for trust, not just capability.

@JayaGup10 Exactly. Decision data is where the real signal lives, not just the final state. If you don’t capture the context, options, and tradeoffs, you lose the “why” behind every action. And without that, teams and agents keep guessing instead of learning.

“Relevant data of course includes the full range of enterprise data sources — structured data, streaming and edge sources, unstructured repositories, imagery data, and more — but it also includes the data that is generated by end users and agents as decisions are being made. This “decision data” contains the context surrounding a given decision, the different options evaluated, and the downstream implications of the committed choice” Welcome

Big companies don’t adopt AI because it’s powerful, they adopt it when it fits how they already work. Consultants bridge that gap, but if context stays fragmented, every rollout becomes another heavy lift. The real shift is when AI carries context across systems so adoption feels lighter, not layered on.

Google is turning consultants into its AI delivery network with a $ 750M fund for firms like McKinsey, Accenture, and Deloitte to help companies build and scale agentic AI. Consulting firms need this because classic consulting work, such as research, slide drafting, process mapping, and software planning, is exactly the kind of work AI systems are starting to automate. AI startups need consultants because big companies rarely buy new tools just because the model is powerful, since they need someone to connect it with data, workflows, security rules, and staff habits. Agentic AI means software that does not only answer questions, but can plan steps, call tools, move through business systems, and complete tasks with less human steering. So Google’s bet is that McKinsey can find the business problem, Google can provide the AI stack, and the client can turn a pilot into a working system across teams. OpenAI’s reported push to sell Codex through Accenture, Capgemini, and PwC points to the same shift, where AI coding tools become enterprise software only after consultants package them into training, governance, and rollout plans. --- businessinsider. com/consulting-mckinsey-accenture-bcg-ai-silicon-valley-enterprise-partnerships-2026-4

Strong point. Speeding up one developer is useful, but most of the pain sits in coordination, not typing. If context and decisions still live in silos, faster code just creates faster misalignment. The real win is making that shared context visible so teams stop relearning the same things every cycle.

every AI coding tool on the market is a single-player game. cursor, copilot, claude code. they all make the individual developer faster at writing code. And they are brilliant at it. 2-5x individual velocity, sometimes more. but software isn't a single-player game. software is mostly architecture decisions, requirements debates, compliance reviews, code reviews, rollbacks, post-mortems, onboarding. it’s a multiplayer sport with a dozen roles and thousands of decisions that don't just live in one person's head. the key in good software development is the coordination between roles, the traceability of decisions, the institutional memory of why something was built a certain way. that's what software factory is. it’s a multiplayer AI. Requirements captures business intent before an engineer opens an IDE. Blueprints captures architecture decisions upstream. Work Orders routes structured tasks to AI agents through MCP with full context. the Knowledge Graph holds the state of every artifact. try it here: factory.8090.ai

Work is changing. Fast. And the next generation of engineers won't just use AI – they'll shape what it becomes. That's exactly what we set out to show at the University of Cebu ICT Congress 2026, speaking to ~2,000 aspiring engineers – one of the biggest college audiences DevRev has engaged with in the Philippines. We put Computer, by DevRev, on stage and let it do the work – live, in real time. Because that's what AI-native engineering actually looks like. But here's the part that matters most: AI can generate. It's engineers who think, validate, and build what actually matters. Anyone can generate code and not everyone can build something that matters. And the engineers who understand that difference are the ones who'll define what comes next. Grateful to the team who brought this to life: Rodolfo Nax Nacu Jr., Arlie Baclayon, Jhoanna Rica Lagumbay, Mira Mohammad, and Jhon Mark Acopiado. What's the one skill you'd bet on for engineers in the age of AI? Comment below!

@tengyanAI Compression without shared context just shifts the load, not the work. Teams should understand that the goal isn’t to work faster, it’s to work softer.

didn’t expect this to hit such a nerve! imo AI agents don’t reduce work as much as they compress it. less typing, more judgment. less waiting, more context switching. less execution, more supervision. that changes how teams should think about productivity, burnout... and what “high output” actually means.

something i've noticed: AI agents create a weird new kind of burnout. esp for young people. a lot of ambitious 22 year olds are going to think the answer is simple: - spin up more agents - ship more code - sleep less - outwork everyone and for a while, it will feel incredible. you can keep multiple agents running, feed them tasks, review outputs, fix mistakes, make decisions, and keep the whole loop moving. the problem is that the work no longer drains you through typing. it drains you through judgment. More attention. More context switching. More verification. More decisions per hour. so instead of 8-10 normal productive hours, you might get 4-5 extremely intense hours before your brain is fully cooked. and you feel numb until you sleep properly and reset some of my friends are already burnt out. they don't say it out loud but i can tell. the agent can keep working 24/7. the human still has a hard limit