Sits with your team
Intent agent
Turns “we need a Postgres for payments, highly available, backups kept six months” into a structured request.
Can only write requests
Self-operating. Human-governed. Provable.
The cloud that runs itself
with an AI ops team built in
Not another cloud provider. Every server, database and cluster comes with AI agents that provision, patch, heal and scale it around the clock — under a control plane that never lets them act alone.
- AI agents on call 24/7
- Humans approve anything risky
- Proof for every change
- DeployAgents plan it. The control plane applies it.
- OperateSelf-healing. Always on.
- ApprovePeople sign off on anything risky.
- VerifyEvery action is signed and auditable.
The work your ops team does today — done by agents
Not a regular cloud
Other clouds rent you the parts. This one runs them.
Global clouds hand you building blocks and a bill for the people to run them. hiveSRE Cloud ships with the people already replaced by software you can audit.
A regular cloud provider
hiveSRE Cloud
You rent servers and databases. Running them is your job.
Every service comes with AI agents that run it for you.
Someone on your team gets paged at 3am.
The Operator diagnoses the failure and takes a permitted fix in seconds.
Patching, resizing and backups wait in a ticket queue.
Routine work happens continuously, as reviewed changes.
“AI for ops” either just chats, or gets the keys and hopes.
Agents propose. A control plane checks every change before anything happens.
Change history is scattered across consoles and logs.
Every change carries a signed record anyone can verify.
Your new operations team
Meet the agents that run your cloud
A few specialised agents, each with one narrow job — and one piece of ordinary software that holds all the authority. A hive of workers, one keeper.
Designs the change
Planner
Works out exact sizes, versions and steps, with its reasoning attached, and opens a pull request a person merges.
Can only propose
On call, around the clock
Operator
Watches everything running. When something breaks it explains what happened and asks for a permitted action or writes up a fix.
Cannot log in or change anything
The keeper · not an AI
The control plane
Deterministic software, and the only part of the system that can actually change your infrastructure. Agents hand it proposals; it decides whether they are safe to apply.
- Validates the change and who it belongs to
- Checks limits and whether it's safe right now
- Snapshots anything that can't be undone
- Stops and asks a human before anything destructive
- Applies it, waits for healthy, writes the record
How a change happens
From one sentence to running infrastructure
No tickets, no runbooks, no console clicking. Every step after your request produces a signed record.
- 1Describe what you needYou
In plain language through the Intent agent, or by editing a small file.
- 2A reviewed change appearsPlanner
Concrete sizes, versions and steps, opened as a pull request with the reasoning.
- 3A person merges itYou
Later, routine changes can be merged by policy.
- 4Checked, applied, confirmedControl plane
Snapshot if needed, apply, and wait until the result is healthy.
- 5Watched from then onOperator
Around the clock, with no dashboards for your team to stare at.
- 6Healed when it breaksOperator
Diagnosed, fixed through a permitted action or a new proposal — back through the same loop.
- You“We need a Postgres for payments. Highly available, backups kept six months.”
- Intent agentRequest recorded — postgres · payments · HA · retention 180d
- PlannerOpened PR #212 — 2 × 4 vCPU, 100 GB, standby in a second zone
- j.smithMerged PR #212
- Control planeChecks passed · applied · healthy · record signed
- OperatorPrimary unresponsive → promoted standby in 41s. No one was paged.
6 events · 5 signed records · 0 people paged
Autonomous, not unsupervised
AI that runs everything and holds no keys
Most “AI for ops” hands a model the keys and hopes. That's fine for a demo and unacceptable for a bank. hiveSRE is built the other way round: agents propose, the control plane acts.
- The AI never has the keys
No agent can create, change or delete infrastructure. Only the control plane can, and only after its checks pass.
- Nothing destructive without a person
Deleting data or downsizing waits for a named human. A snapshot is taken first.
- Free text is read, never executed
Agents work from structured intent and produce structured, reviewable changes.
- Models run inside, not in someone's API
Agents use models on your own hardware. No hosted AI APIs, no telemetry, no call-outs.
Record ev‑7f3a2dVerified
- Seen
- Primary unresponsive for 20s
- Proposed by
- Operator
- Action
- Promote standby (permitted)
- Policy
- Production DB failover
- Result
- Healthy in 41s
sig ed25519 · 9c1f…a47e
Every change and every emergency action produces a record like this: what the agent saw, what it proposed, which policy allowed it, who approved it and what was done. Logs can be edited. These can't — and anyone can verify them offline.
Services
Everything a cloud has. Operated by agents.
The building blocks you'd expect from a cloud — each one operated by the agents from day one. Some are still on the roadmap; ask us which.
Compute
Virtual machines on fast local storage, right-sized and patched by agents.
Managed PostgreSQL
High availability, point-in-time restore and replicas. The Operator handles failover.
Kubernetes
Managed clusters with upgrades and node health handled for you.
Object Storage
S3-compatible buckets, with retention locks where you need them.
Networking
Private networks, firewalls and load balancers, kept in line with your policy.
AI Inference
Private language models, including Hebrew-capable ones, that never send data outside.
Agent Sandboxes
Isolated micro-VMs for running your own AI agents safely.
Git
Private hosting for the files that describe your infrastructure. Rollback is usually one revert.
Evidence
Signed records for every operation on your account. Always on.
Where it runs
Same agents, wherever your cloud lives
Operations are included in every option, so there's no ops headcount on top. Nothing inside calls out to the internet — the agents use models running on the same hardware.
Hosted
Region pendingA ready-made cloud with the agents already running it. Start in days.
From ₪0.09per vCPU-hour
- Shared region on Israeli soil
- All services, agents included
- No ops headcount on top
- Signed record for every change
Dedicated
Most popularYour servers, in your data center or colo cage. Our control plane and agents run them.
Customper server, per month
- Single-tenant, your hardware
- Agents and control plane included
- Hardware sourcing available
- Data never mixed with anyone's
Air-gapped
No internet at all, and the agents still run everything.
Customannual license
- No internet connection
- Signed offline update bundles
- Agents use models run locally
- Offline-verifiable evidence
Questions
What people ask about an AI-run cloud
A regular cloud rents you the parts and leaves running them to you. hiveSRE Cloud includes the operations team: AI agents that provision, patch, heal and scale your services around the clock, under a control plane that checks every change.
No. Agents can only propose. The control plane — ordinary, deterministic software, not an AI — is the only thing allowed to change infrastructure, and only after its checks pass. Anything destructive waits for a named person.
A bad proposal fails the control plane's checks or is rejected in review. Snapshots are taken before anything that can't be undone, and because every change lives in git, rolling back is usually a single revert.
Far less of one. Provisioning, patching, failover, tuning and 3am pages are handled by the platform. Your engineers decide what the platform should be and approve anything risky.
No. It works from structured intent and produces structured, reviewable changes. Free text is read, but never executed.
Hosted in Israel, on your own hardware, or fully air-gapped. The agents use models running inside that environment; nothing calls out to the internet.
A signed evidence record for every change: what the agent saw, what it proposed, which policy allowed it, who approved it and what was done. It can be verified offline, without access to our systems.
Design partners
Hand the pager to agents. Keep the final say.
We're onboarding a small number of design partners who will shape the roadmap and work directly with the founding team. Tell us what you'd like the agents to take off your plate.
Prefer email? [email protected]