Plural helps platform teams get to Backlog Zero with governed AI agents.
They run inside your environment and take cluster upgrades, incident triage and CVE patching end to end. Every change comes back as a PR you review.
#MSIgnite SF · Booth #5135 · Nov 17–20
2
87
Some teams at KubeCon have AI agents handling upgrades, incident triage and CVE patching. Others use them as a copilot.
Our guide, Evaluating Infrastructure Autonomy, maps 5 stages to help you plan your next step toward agents in production.
Download: dub.sh/Q4DQpmD
1
112
#KubeCon is all about agent guardrails, identity, sandboxes and gateways. Teams are still figuring out how to trust agents with infra.
Plural grounds agents in your whole fleet, and every upgrade, CVE or ticket they take on lands as a PR you review.
Backlog Zero at booth #192.
2
125
We wrapped up October with a double-speaker-meetup featuring @clesleycode with her talk on data governance for observability, and @michaeljguarino sharing his work in rebuilding loki, prometheus, and tempo in rust!
Thank you so much to @TeraSkycom1 for hosting our evening - your decor is on point. 🙏
📍 Over the next few weeks, the Plural team is on the road and will be making an appearance at: @KubeCon_ SLC, Microsoft Ignite, @AWSreInvent, and Gartner IT IOCS. If you're headed to any of these conferences, let us know! We'd love to continue (or start) the conversation there.
We will have one final meetup this year, taking place in December. Stay tuned for updates! See you soon 👋
1
124
Plural.sh retweeted
Proud that @Microsoft has named @plural_sh as one of 14 startups to watch at #Ignite. Our governed AI agents clear CVEs, incidents and K8s upgrades, with every fix as a reviewable PR.
Come see it live at Innovation Basecamp in SF, Nov 16-18.
microsoft.com/en-us/startups…
Every platform team has a backlog of upgrades, CVEs, tickets and drift that never gets done.
Plural’s AI agents run in your environment and take that work end to end.
What’s at the top of yours?
Talk to us about getting to Backlog Zero at #KubeCon SLC · Booth #192 · Nov 9–12
1
1
116
Plural.sh retweeted
This is why Plural built its agents to be cloud only from the start. We interact with platform-level permissions, it's a serious responsibility and requires control.
Plural's GitOps engine also enables a world where eng laptops can be cleaned of cloud credentials entirely too.
📣 Last call for the October Plural x Kubernetes meetup happening on Thursday! Seats are very limited so please be sure to RSVP ASAP at luma.com/05srus23
You will get to hear from Lesley Cordero on data governance for observability, and Michael Guarino on lessons from rewriting Prometheus / Loki / Tempo in Rust. Plus, you can meet peers working in the DevOps space! We will have food and drinks as usual~
See you soon!
2
129
Join us on Thursday, 10/29 at 6pm for the Plural x Kubernetes October meetup 👋 We have a double header event once again!
💡 Our first guest speaker is Lesley Cordero, Staff Engineer & Tech Lead at The New York Times. Her session topic is "Who Watches the Watchers? Data Governance for Observability."
💡 Our second speaker is Michael Guarino, CTO at Plural. He will share lessons from rewriting Prometheus / Loki / Tempo in Rust.
A BIG thank you to our friends at TeraSky for partnering to make this meetup possible 🙏
➡️ Seats are limited, save yours at: luma.com/05srus23
2
143
What a highlight last week 🌟 Thank you all for joining our September meetup and asking (many!) questions to our two speakers. Key takeaways:
Neil Carpenter spoke about how vulnerability management in containerized workloads can be painful. Security teams are implementing vuln mgmt primarily to address risk concerns and compliance concerns. Three broad approaches to make this better:
⬆️ Prioritize: Use vuln metadata (how severe / how exploitable), asset context (is it exposed / is it important / does it get me to something else important), and reachability (can this vuln actually be reached in our deployment?)
⬇️ Minimize: Reduce excess packages in base image and codebase to have fewer vulns.
🔄 Automate: Where possible, automate patching and updating base images to address new vulns faster.
Zlatko Lakisic demo'ed running AI agents on Kubernetes, sharing his open source engine that turns plain language goal into execution via agentic tools:
🧩 Each planned step is a unit of work. The planner reads the results and revises the plan before the next step.
⏱️ Measure the overhead: starting an empty worker Job took roughly 3 seconds in his tests, even with the image already on the node. His edge deployment uses long-lived worker pods to avoid repeating startup cost for every step.
🔀 Match execution to the task: use a fresh Job for isolation, a warm worker pool when latency matters more, or run within the engine process when a step needs the client’s local tools.
🔐 Clients bring their own agents and tools and connect through a single socket. They never receive cluster credentials - only the coordinator can create Jobs.
A BIG thank you to our friends at @TeraSkycom1 for hosting us in your lovely space, and for being so engaged throughout the evening 🫶
If you'd like to speak or demo at our next meetup, please DM or comment 👋 We will see you in October!
2
159
📣 Last call for the Plural x Kubernetes September meetup tomorrow! We have a double header event 🎤🎤 Also, thank you to our friends at @TeraSkycom1 for partnering to make this meetup possible 🙏
✅ RSVP ASAP at: luma.com/qr8xi06s
🗓️ Schedule:
6:00pm - door opens
6:30pm - intros (please arrive by this time!)
6:40pm - speaker programming
7:20pm - networking
8:00pm - event ends
Our first guest speaker is Neil Carpenter, Field CTO at large. He will be exploring the question: What if we just stopped patching our container images?
Session Abstract: The security team wants to shift left, which sounds a lot like moving additional work into your day. In this presentation, Neil will dive into what security teams are actually hoping to accomplish, address why initial implementations tend to cause more pain and worse outcomes, and sketch out strategies for fixing that.
Our second speaker is Zlatko Lakisic, Senior Architect, Emerging Technology & AI, at New York Power Authority. He will be demo-ing running AI agents as Kubernetes jobs.
Session Abstract: Agentic Orchestration is an open source engine I built that turns a goal into a plan and runs each step as an agent. On Kubernetes, every step runs as a Job in its own pod, and the coordinator only plans and dispatches. I run it on a single node k3s cluster at the edge, and we will cover what pod per step actually costs. For the demo, I will connect voice assistant COMSTAR which has its own agents and tools. The infrastructure is created when it connects and we will watch the jobs appear and finish in kubectl. Voice in, pods out.
We will have food and drinks during this event. Please arrive no later than 6:30pm so we can get started promptly!
3
176
Plural.sh retweeted
Replying to @brexton
This is how we've enabled F500s to use AI at Plural. Agents hosted in their infrastructure, orchestration handled for them, coding agents running in restricted pods without production credentials, and infra access governed by their RBAC and OPA policies (often a big blocker).
We are back with a double header for the September Plural x Kubernetes NYC meetup! Join us next Thursday, 9/24 and you'll get to hear from Neil Carpenter and Zlatko Lakisic 👋
✅ RSVP at luma.com/qr8xi06s
💡 Neil will be exploring the question: What if we just stopped patching our container images?
Session Abstract: The security team wants to shift left, which sounds a lot like moving additional work into your day. In this presentation, Neil will dive into what security teams are actually hoping to accomplish, address why initial implementations tend to cause more pain and worse outcomes, and sketch out strategies for fixing that.
🤖 Zlatko will demo running AI agents as Kubernetes jobs.
Session Abstract: Agentic Orchestration is an open source engine I built that turns a goal into a plan and runs each step as an agent. On Kubernetes, every step runs as a Job in its own pod, and the coordinator only plans and dispatches. I run it on a single node k3s cluster at the edge, and we will cover what pod per step actually costs. For the demo, I will connect voice assistant COMSTAR which has its own agents and tools. The infrastructure is created when it connects and we will watch the jobs appear and finish in kubectl. Voice in, pods out.
🗓️ Schedule:
6:00pm - door opens
6:30pm - intros (please arrive by this time!)
6:40pm - speaker programming
7:20pm - networking
8:00pm - event ends
See you there!
4
256
Neocloud infrastructure is inherently fragmented across separate clouds, data centers, and hundreds of clusters, making it difficult to know what is vulnerable and remediate it consistently.
Plural provides a single pane of glass across all your clusters and uses enterprise-governed agents to remediate CVEs end-to-end. Check us out.
Neoclouds have limited cybersecurity. Next time agents successfully go rouge, they'll try taking over a neocloud to run more copies. This is bad.
Thus: neoclouds should greatly strengthen their cybersecurity and every company with strong cyber models should help with that.
3
498
We're at KCD San Francisco today! Come say hi and learn about Plural's self-driving K8s agents. #KCD #CNCF #CloudNative
As the article puts it, neoclouds need “a SYSTEM to get stuff patched”.
Neocloud infrastructure is inherently fragmented across separate clouds, data centers, and hundreds of clusters, making it difficult to know what is vulnerable and remediate it consistently.
Plural provides that system, a single pane of glass across the entire fleet, and uses governed agents to remediate CVEs end to end.
Most Neoclouds Suck At Security
OpenAI vs HuggingFace, Container Escapes,
Kernel Bypass, Network Policies, Security Keys,
Multi-tenant Grafana, and a ClusterMAX 3.0 Preview
semianalysis.substack.com/p/…
3
325
Our special meetup with @elastic took place in their beautiful office, with two speaker sessions that's standing room only! Key takeaways from @michaeljguarino & Eric Mustin's talks 👇
📦 The traditional observability stack bundles telemetry storage, collection, dashboards, and alerting into one platform. That convenience has historically given all-in-one vendors significant pricing power.
📤 That stack is becoming easier to disaggregate. OpenTelemetry is maturing as a vendor-neutral collection layer, while specialized open-source tools can handle the underlying data.
🤖 AI agents can make this modular stack easier to operate. They can help manage complex datastores and perform analyses on demand, allowing teams to focus more of their observability spend on retaining and processing high-quality data.
🔄 The principle is “automation first, agents second.” Practices like automated index rollover, balanced shard sizing, Kubernetes operators, and GitOps provide the guardrails agents need to make infrastructure changes safely.
⛵ Telemetry takes a perilous journey from the Pod to a Dashboard. It can get lost, changed, and managed at many stops along the way.
🔍 Prometheus and OpenTelemetry Operators are easy ways to collect Monitoring and Observability Telemetry from K8s, without code changes.
🚒 Telemetry collection is a firehose out of the box. Cost efficiencies are possible with best practice normalization, filtering, processing, and sampling in your telemetry collection pipeline.
A BIG thank you to Jonathan Avezbaki, Olivia Petrie, and Joy Jing for making this long-awaited collab possible 🙏
Our next meetup is in late September - stay tuned! Leave us a reply if you're interested in presenting at a future event :)
4
221
📣 One week away: Plural x Elastic x Kubernetes August meetup with speakers @michaeljguarino & Eric Mustin!
Seat are limited - RSVP ASAP if you can make it 👉 luma.com/o9zifks0
💡 Michael's topic is "Elasticsearch and Agentic Ops: A Perfect Match" 💡
Michael's talk discusses the ways we've used Elasticsearch at Plural to implement e2e agentic SRE workflows. Examples include using agents to manage your ES clusters themselves, to leveraging it as a low cost log storage engine for deep application root cause analysis, enabling you to reduce observability cost without sacrificing developer experience while maintaining realistic operational total cost of ownership.
💡 Eric's topic is "Telemetry’s Long Way Home: A Kubernetes Odyssey with Prometheus, OpenTelemetry, and Elastic" 💡
Eric's talk explores how Prometheus, OpenTelemetry, and Elastic can work together to give Kubernetes metrics, logs, and traces a reliable path home. We’ll look at practical collection architectures, Prometheus-native metrics, OpenTelemetry processing and routing, and how Elastic can bring the resulting signals together. We’ll also discuss resiliency, duplication, architectural tradeoffs, and how to avoid getting lost at sea in an increasingly complex observability ecosystem.
See you soon!
2
155


















