MetalBear

monday.com has 2,000+ AI agents in production, merging thousands of PRs. The agents pick up tasks off the same monday boards as the engineers, write code and test their changes against real services in Kubernetes with mirrord. Then they open PRs that have already been fully verified, ready for an engineer to review. So what would it take to build this for your team? On Nov 4, join Uzi Abdullayev, Senior Software Engineer | DevEx AI, and Tomer Brook, Senior AI Engineering Manager, from monday.com, with host Eyal Bukchin, CTO at MetalBear, and learn: → how a task moves from the board to a tested PR → how agents hand work off to each other → what they learned scaling it, and where you can start Bring your questions. We'll answer them live. 📅Wed, Nov 4 · 10am ET / 4pm CET Save your spot 👇 https://lnkd.in/duvHCDXQ Can’t make it? Register anyway and we’ll send you the recording. #AgenticAI #PlatformEngineering #Kubernetes

A new model or a prompt tweak can quietly break the agent your customers rely on. See how monday.com catches it. On demand.

Stop spinning up a staging copy for every agent evals run. In 30 min, our engineers show you how to use the cluster you have.

What do your agent evals actually run against? Mocks, or the real services your agent calls? Full session, on demand.

Most teams eval their AI agents against mocks and hope for the best in production. At monday.com, we stopped hoping. We built an eval framework that runs our agents against real staging services and real data, the same dependencies they'll hit in production. The result: agent evals became a merge gate in CI. A model change, a prompt update, a new tool call: if it breaks something real, we catch it before it ships. On September 15th, I'm joining Eyal Bukchin, Co-founder & CTO of MetalBear, to walk through the full setup live. We'll cover the eval suite, how mirrord connects agents to staging dependencies, and how to wire it all into your CI pipeline. If you're building or shipping AI agents at scale, this one's worth an hour of your time. Free to attend. Link: https://lnkd.in/d3-PySaF

How monday.com Runs Agent Evals Against Real Dependencies A prompt tweak, a model upgrade, or one changed tool can quietly break what your customers actually experience. That risk grows when an agentic product spends its time calling databases, internal services, and third-party APIs. Evals are how you catch it before customers do. But the suite still needs something real to test. Stand-ins can drift from the integrations the agent actually calls, while spinning up a full environment for every run costs time and money. monday.com took a third route. Their eval suite runs against their staging cluster, with the real services, populated databases, and queues, on the pull request and before the merge. 📅 On Tuesday, September 15, join Eyal Bukchin (Co-founder & CTO, MetalBear) in conversation with Dor Cohen (AI Engineering Director, monday.com) for a live interview, two live demos, and audience Q&A. What you'll see - The design decisions behind monday.com's agent eval setup - A mirrord demo wiring an eval gate to a cluster from scratch - monday.com's own pipeline running on a real PR What you'll walk away with - The full path: eval suite → mirrord → staging services and data → merge gate in CI - An example repo with the eval suite and the CI config, plus a setup guide 📅 Tuesday, September 15 · 10am ET / 4pm CET · 45 minutes including Q&A · Free, live, and recorded ⚠️ Register here: https://lnkd.in/dh9F7zPU LinkedIn’s “Attend” button only marks your interest in the event. Register through Zoom above to secure your place and receive access to the session.⚠️ #AIAgents #Kubernetes #PlatformEngineering #DevEx #AgentEvals #mirrord #MetalBear