The shift toward specialized AI agents continues gaining momentum, with security taking center stage as developers demand tools that actually understand their workflows. As we’ve been tracking over recent weeks, generic AI assistants are steadily giving way to purpose-built agents that integrate seamlessly into existing development processes.
What’s particularly striking this week is how testing and quality assurance are evolving alongside these agent deployments. Teams are moving beyond reactive debugging toward proactive systems that catch issues before they impact users.
The funding activity reflects this ongoing maturation, with investors backing solutions that solve specific, measurable problems.
🏆 Funding Wins
Qbeast Bags $7.6M To Banish Data Bottlenecks
Barcelona-based Qbeast raised $7.6M in seed funding led by Peak XV, with HWK TechInvestment and Elaia Partners joining the round. The Barcelona Supercomputing Center spinout tackles a massive problem: up to 90% of compute resources get wasted scanning irrelevant data in modern data lakes.
Founded by Cesare Cugnasco and Paola Pardo, with new CEO Srikanth Satya (former AWS and Azure veteran) now leading the charge, Qbeast plugs directly into Delta Lake, Apache Iceberg, and Hudi tables with multi-dimensional indexing that handles complex filters across time, region, and customer segments simultaneously. The platform delivers 2-6x query speedups and 70% cost reductions without vendor lock-in.
🎯 Congratulations to the team! Read the full funding announcement here.
CodeKarma’s $2.5M Brings Production Insights To Dev Workflow
Bengaluru startup CodeKarma secured $2.5M in pre-seed funding from Prosus, Accel, and Xeed Ventures, with SenseAI Ventures and Stargazer Ventures participating. The company made this announcement just 3 weeks after launching!
Co-founders Anantharam Vanchi Prakash and Priyanka Nahata built the platform to embed real-time production insights directly into developer workflows. The AI-native infrastructure enables risk detection, dead code removal, and continuous software quality improvement through closed-loop feedback systems. The funding will drive US enterprise expansion and product development for managing large, evolving codebases. CodeKarma serves engineering teams across e-commerce, fintech, and logistics, helping them streamline operations and reduce technical debt.
🔥 Well done on the rapid traction! Check out CodeKarma for smarter development workflows.
$20M For Tavily’s AI Agent Safety Controls
New York’s Tavily raised $20M Series A co-led by Insight Partners and Alpha Wave Global, bringing total funding to $25M. The one-year-old startup helps companies connect AI agents to the internet while enforcing corporate compliance guardrails.
Founded by data scientist Rotem Weiss, Tavily evolved from his viral open-source project GPT Researcher, which gained 20K GitHub stars. The platform provides search, crawl, and extraction tools for enterprise clients like Groq, Cohere, MongoDB, and Writer, ensuring agents access real-time web data without creating compliance nightmares.
AI agents now handle fraud detection in finance and customer research in sales, but connecting them directly to the web without safeguards creates a “wild west” scenario. Tavily’s governance layer solves this problem while the company aims to onboard the next billion agents to the web safely.
⚡ Impressive growth trajectory! Explore Tavily for compliant agent web connectivity.
🚀 3, 2, 1… Launches
Asteroid Opens Browser Agent Platform After YC Success
Asteroid (YC W25) opened its browser agent platform to the public after six months of battle-testing with startups and enterprises. The YC February cohort startup, founded by David Mlčoch and Joe Hewett, initially focused on back-office automation before deciding to democratize their technology.
The platform takes a different approach from “do-everything” agents that frequently break. Instead, Asteroid enables users to build hundreds of specialized, reliable agents in minutes without coding. Domain experts like insurance brokers can create their own automation rather than explaining every click to developers. Real production use cases include insurance quoting across multiple providers, medical appointment data entry, invoice retrieval from supplier portals, and voice agent integration with legacy systems.
🤖 Smart specialization strategy! Build browser agents with Asteroid and try their AGI Escape Room challenge. Show your support on Product Hunt too!
Apiiro Announces AutoFix Agent For AppSec Teams
Apiiro launched AutoFix Agent, an AI assistant delivering automated fixes to design and code risks with single-click remediation. The platform tackles a critical problem: 50% of AI-generated code contains vulnerabilities, with 10% actively exploitable, while integrating seamlessly into developer IDEs via MCP connection.
Co-founder Idan Plotnik also revealed their ServiceNow partnership last week, explaining why ServiceNow called it an ‘industry-first’:
“Apiiro is the only company that can automatically discover, inventory, and visualize all Code Assets across millions of codebases in Fortune 500 companies, map their relationships into a software architecture graph, and match them to runtime deployments.”
🤖 Impressive work Apiiro! Try AutoFix Agent for smarter AppSec workflows and if you’re an AppSec Engineer looking for a new role, send them a message.
OX Security Announces Agent OX For Custom Code Fixes
In another AI-powered AppSec launch this week, OX Security introduced Agent OX, an assistant that automatically generates organization-specific code fixes beyond just identifying vulnerabilities. Co-founder and CEO Neatsun Ziv explains their approach:
“The new system is not a generic recommendation. It uses the developer’s own writing style and the names of the parameters, and the context used in the ecosystem.”
Agent OX uses multiple AI agents representing different viewpoints like architects and security specialists to analyze code architecture and runtime context. Developers can then review the generated fixes and approve them with one click.
⚡ Explore Agent OX for tailored vulnerability fixes.
SignalWire Opens Call Fabric SDK For Modern Communications
SignalWire announced the open beta of its Call Fabric SDK and Reference App, empowering developers to build full-stack communications apps across web, iOS, and Android. The JavaScript-based SDK includes React Native adaptors for mobile development. CEO and co-founder Anthony Minessale II explains the mission:
“SignalWire was founded to put Programmable Unified Communications in the hands of developers. With our new SDK and Reference App, we’re removing barriers and giving builders a truly open, flexible platform to bring advanced communications and Voice AI solutions to life on their terms.”
The open-source Reference App provides a foundation for everything from click-to-call widgets to full UCaaS applications.
📞 Excellent timing for communications innovation! Start building with SignalWire and join their developer community.
AgentHub Launches Realistic AI Agent Testing Platform
Y Combinator S25 startup AgentHub debuted as an implementation-agnostic platform for evaluating AI agents in realistic sandbox environments. Founded by Youssef Kallel (former Apple Foundation Model Evaluation team lead) and Sandra Tang (MIT CS graduate with experience at Google and Meta).
LLM-based agents often pass simple evaluations but fail in deployment through hallucinated tool use and inefficient reasoning. AgentHub provides customizable sandboxes for e-commerce, CRM, browsers, and file systems, with detailed step-level traces capturing every LLM call and UI interaction. The platform offers plug-and-play evaluation with LLM, rule-based, or human-in-the-loop grading at task, step, and tool levels.
🧪 Are you building agents or training models? Test with AgentHub before shipping to production, and reach out to the team.
Synthetic Society Brings AI-Powered User Testing To Market
Synthetic Society (YC S25) launched synthetic user testing to catch bugs, UX flaws, and broken flows before real users encounter them. Founded by Aaron Chew and Kavan Doctor, competitive programmers who dropped out of Columbia and MIT to build the platform.
The startup deploys swarms of AI-driven synthetic users that mimic real user behaviors, proactively identifying issues in an era where AI code generation accelerates feature shipping. Teams can eliminate manual testing, accelerate release cycles, and ship with confidence using self-updating end-to-end flows that protect against changes.
🔍 Brilliant approach to quality assurance! Try Synthetic Society for proactive bug detection.
Snyk Unveils ‘Secure At Inception’ For AI-Native Development
Snyk announced Secure At Inception, three innovations focused on Model Context Protocol technology for AI coding assistants. The release includes deeply-integrated real-time security scanning, visibility into generative AI components, and an experimental scanner for AI-specific MCP vulnerabilities.
The platform embeds security into agentic workflows as developers shift to “vibe coding” with high-level prompts instead of traditional code, addressing the expanded risk surface of AI agents interacting with external tools and data.
🛡️ Essential for AI-native security! Explore Secure At Inception for embedded AI development protection.
Mistral AI Ships Codestral 25.08 With 30% Better Performance
MISTRAL AI rolled out Codestral 25.08 alongside their complete enterprise coding stack, delivering 30% more accepted completions, 10% more retained code, and 50% fewer runaway generations. The update includes Codestral Embed for codebase-scale search and Devstral for autonomous multi-step development.
The integrated platform addresses enterprise limitations like deployment constraints, limited customization, and fragmented architecture. Leading organizations including Capgemini, Abanca, and SNCF are already using the stack in production across consulting, finance, and transportation sectors with full data control and compliance requirements met through on-premises deployment options.
💻 Great work on the enterprise focus! Try Codestral 25.08 for improved coding performance.
🔦 DevTools of the Week: VibeFlow
VibeFlow (YC S25) represents what happens when two brilliant minds refuse to settle for “good enough.” Co-founders Alessia Paccagnella and Elia Saquand found each other after months of searching YC’s co-founder matching platform, realizing they were both “complaining” about not finding people with matching ambition and drive.
Alessia left her “perfect job” in Zurich, while Elia was grinding through Entrepreneur First. Both ETH Zurich graduates with impressive backgrounds – Alessia led the design and development of a new visual programming tool at Esri, Elia accelerated AlphaFold training by 10x at InstaDeep and built perception systems at Sony AI – they spent months pivoting through biotech, dermatology clinics, and influencer marketing before finding their calling.
The breakthrough came when they focused on AI and dev tools. VibeFlow solves what every builder faces: AI tools excel at creating frontends you can see and iterate on, but backends remain impenetrable black boxes. When something breaks, you’re stuck re-prompting and hoping for the best.
VibeFlow changes this by generating full-stack web apps with n8n-style visual workflows for backend logic. Think “Lovable meets n8n” – you get the rapid frontend generation but with complete visibility and control over your backend architecture, integrations, and data flows.
After multiple YC rejections, they made it into S25 with just 0.6% acceptance rate. As Alessia puts it:
“If you’re going to do something, do it FULLY. No half-measures, no safety nets.”
🪽 Brilliant execution on a universal problem! We’re fully behind this team! Build with VibeFlow and show them some love 👏
💰 New Funds for Founders
SIVentures (prev. Smart Infrastructure Ventures) launched a €30 million second fund targeting 30 early-stage B2B startups across AI, software, and enterprise tech in Germany, Austria, and Switzerland. The Leipzig-based firm has been backing DACH region founders for six years. Find out more here
Airtree closed a massive $420 million fifth fund with backing from Harvard, MetLife, and the University of Wisconsin. The 11-year-old Sydney firm’s fundraise signals growing foreign interest in Australia’s tech scene, which has produced unicorns like Canva and is finally getting the global recognition it deserves. Find out more here
Got a launch, raise, or tool we should feature next week?
Reach out to one of our team.
Louise Ogilvy (Recruitment Director) – UK/USA
Becca Combe MIRP CertRP (Lead Recruiter) – USA
Ben Chipchase (Lead Recruiter) – UK
Natalie Harper (Marketing Administrator)
Looking for your next DevTools role?
Visit devtoolsjobs.com
