The AI-native operating system for modern DevOps teams.

It understands your stack, runs the work, and proves the outcome.

The problem

Teams don't lack DevOps tools. They drown in them.

Kubernetes

GitHub

Terraform

ArgoCD

AWS

Datadog

Grafana

PagerDuty

Vault

Slack

Jira

Coralogix

Cloudflare

Snyk

GitLab + 30 more

Every tool ships its own AI None of them share context The engineer is the integration layer

What we build

One platform that understands your whole DevOps environment.

LAYER 01

Unified Context

Every source across the stack, joined into one model.

LAYER 02

AI DevOps Team

Specialists on one shared context and one shared memory, led by Ezra.

LAYER 03

Operational Intelligence

Finds risk, toil and waste before anyone opens a ticket — then acts and measures.

Layer 01 · Unified Context

For the first time, one source of truth.

InfrastructureK8s · AWS · Terraform

CI / CDActions · Argo · GitLab

ObservabilityDatadog · Grafana

Security & AccessVault · Snyk · IAM

FinOpsCloud billing

IncidentsPagerDuty · Jira

Runbooks & DocsConfluence · Notion

Team & CommsSlack · ownership

The Unified Context Graph

Entities services, clusters, pipelines, people

Relations what depends on what, who owns it

History every change, incident, and cost curve

Memory decisions and fixes that stay learned

Layer 02 · AI DevOps Team

A team, not a chatbot. And someone runs it.

Orchestrator

Ezra

Senior DevOps Engineer

Plans. Turns a goal into a concrete plan of work.

Delegates. Routes each task to the specialist who owns it.

Verifies. Checks the result and reports what changed.

Kubernetes Expertcluster · workloads · drift

Incident Commanderdetection · triage · MTTR

Security Engineerposture · access · exposure

Platform EngineerIaC · pipelines · golden paths

FinOps Analystspend · anomalies · waste

Compliance Officercontrols · audit · policy

One context · one memory · one owner of the outcome.

Layer 03 · Operational Intelligence

It doesn't wait to be asked.

Reliability

Bottlenecks and load hotspots

Recurring failures

Unstable services

Risky deployments

Cost

Cost anomalies

Idle and over-provisioned resources

Spend with no owner

Security & Drift

Kubernetes drift

Broken permissions

Exposure and posture gaps

Organization

Missing documentation

Areas with no owner

Unbalanced team workload

Autonomy

Hand over the toil, not the keys.

AUTONOMOUS

Investigate & recommend. Correlate signals, draft the postmortem, open the ticket, propose the fix, watch for regressions.

NEEDS APPROVAL

Change the running system. Roll back a deploy, scale a resource, rotate a credential.

NEVER

Cross the hard lines. Destructive, irreversible, or policy-bound actions stay with people.

The real differentiator

We don't just run AI. We measure the DevOps operation.

MTTR

time to recover

CFR

change failure rate

AI %

tickets closed autonomously

Toil ↓

repetitive work removed

Why does this service always fall over?

Who owns Production Payments?

Which deployment spiked latency last week?

How much time did Ezra save us this month?

The shift

Same stack. A fundamentally different way to operate it.

Today

Operating Platform

— A dozen dashboards, each a silo

+ One shared context across the stack

— An AI bolted onto every tool

+ One AI DevOps team that coordinates

— Knowledge trapped in one engineer's head

+ Shared organizational memory

— Observe and alert, then page a human

+ Observe, operate, and act within limits

— Context resets every session

+ Gets measurably better over time

Market

A new category, forming where three big markets collide.

DevOps tooling, observability & AIOps, and platform engineering — converging into one operating layer. Figures directional

TAM

$50B+

Global DevOps, observability and platform tooling · growing ~20% a year.

SAM

~$12B

Mid-market & enterprise orgs adopting AI-native operations.

SOM · entry wedge

Design partners → early enterprise

Land in cloud-native teams; expand across domains and seats.

LLM reasoningAgents + MCP Cloud maturityTool sprawl Do more with lessAI-native adoption

Competitive landscape

A market of specialists. One missing operating layer.

AI-native, narrow The operating layer Point tools Broad, but not autonomous Point solution Whole DevOps operation AI team + shared memory Single agent / no agent Us

Claude

Viktor

Resolve.ai

Cleric TalkOps Monk

Sedai

Harness

Komodor

GitLab

Datadog

PagerDuty

Dynatrace

Us AI SRE & DevOps agents Horizontal AI agents Incumbent platforms & observability Our read of positioning

The incumbents

Huge surface, deep data — but they observe and alert. No team, no memory, no ownership.

The AI wave

Real intelligence, one agent, one lane — and it forgets between sessions.

The open corner

Nobody pairs whole-operation coverage with a coordinated AI team, shared memory and safe autonomy.

Capabilities

What we have that nobody else puts in one product.

	Us	Incumbent platforms	AI SRE specialists	Horizontal AI agents
Whole DevOps operation	✓All domains	~Broad, but one discipline each	✕One lane	~Any tool, no depth
A team of specialists	✓7 roles, one orchestrator	✕No agents	✕Single agent	~Generic sub-agents
One owner of the outcome	✓Ezra orchestrates	✕The engineer is the glue	✕Task-scoped only	✕Prompt-scoped only
Shared operational memory	✓Compounds per customer	✕Dashboards, not memory	~Per-incident recall	✕Resets each session
Safe autonomy on production	✓3-tier, policy-enforced	~Scripted automation	~Read-mostly	✕No production guardrails
Proactive, not prompted	✓Always running	~Alerts on thresholds	~Triggered by incidents	✕Waits for a prompt
Measures the operation	✓MTTR · CFR · AI% · toil	~System metrics only	✕Not measured	✕Not measured

Any one row is buildable. All seven in one system is not — they only work when they share the same context and the same memory.

Build vs. buy

"Can't a team just build Ezra?"

They can start. Here's the bill.

Build it yourself

Tool integrations~40 APIs, each with its own auth & rate limits

Entity resolutionSame service, five different names

Shared memoryStorage, retrieval, decay, per-tenant isolation

Agent orchestrationPlanning, routing, retries, long-running state

Safe autonomy & guardrailsBlast radius, dry runs, rollback, kill switch

Policy & auditWho approved what, and can you prove it

Evals & QAProving the agent didn't get worse this week

Forever upkeepEvery API change, model change, new tool your team

8 systems · ~20 vendors · and your team owns every one, forever.

vs

Buy the platform

One operating platform

Ezra and the AI team, unified context, shared memory, and safe autonomy — one system, not eight.

Build 12–18 months · 3–5 engineers

Buy Connected in days

And the memory keeps compounding on our side, not on your backlog.

Viktor and Claude hand you a raw agent — you'd still build everything on the left. AWS existed. Vercel still won.

Why us

We've spent our careers inside this problem.

Operators first

We've written Terraform at 2am. We didn't read about this problem.

Pattern recognition

After enough companies, you stop seeing unique problems. You start seeing the same broken operating model.

Built for production

Cybersecurity taught us one thing: autonomy only matters when it's trusted, auditable, and safe.

We're building the platform we always wished we had.

The founders

Built for DevOps, by DevOps.

LK

Liraz Krispel

Co-founder

Sola SecurityPlatform Engineering Team Lead
Palo Alto NetworksStaff DevOps Engineer
Cider SecurityDevSecOps Engineer
Office of the Prime MinisterNetwork & Security Specialist
Unit 2800 (IDF)Transmission & Networking Engineer

SV

Sergey Vinogradov

Co-founder

ParagonDirector, Delivery & DevOps
Office of the Prime MinisterIT Specialist

DL

Dvir Levy

Co-founder

PenteraCyber Research Team Lead
RiskifiedSecDevOps Engineer
Comm-ITAWS Cloud Architect & DevOps
Unit 2800 (IDF)System Administrator · Team Lead Asst.

OA

Omer Aviram

Co-founder

SimilarwebSenior Product Designer & DS Lead
KELA CyberProduct Designer

LM

Liad Mazar

Co-founder

BlinkOpsDevOps Engineer
IdomooIT Manager
Israel PoliceDetective

The ask

Raising to prove the wedge and earn the next round.

The raise

$8–10M

Seed · 18-month runway

Product & engineering60%

Design partners & go-to-market25%

Operations & G&A15%

Now · 0–3 mo

Land design partners. Read-only context + the cost & ownership map live.

6 mo

First killer outcome in prod. Cost tied to deploy and owner; Ezra recommends fixes.

12 mo

Paying partners + approved actions. Expand to a second domain; early revenue.

18 mo

Repeatable ICP & ARR. Proven wedge and expansion, ready to raise the A.

We're building the operating system for AI-native DevOps teams.

Not a tool that runs commands — a system that runs the operation.