Developer speed.
Enterprise control.
One platform.
Move your applications from Git to a secure Kubernetes production environment. Manage build, security scanning, deploy, identity, observability, backup and AI cost management in a shared control plane.
Manage build, security, identity, operations and AI cost on one platform instead of scattered tools.
All your tenants, projects and services in a single console; no more getting lost across scattered panels and tabs.
An automatic chain from Git repository to HTTPS address; no waiting for a per-app DNS record or infrastructure ticket.
Every build goes through secret, code, dependency, image and SBOM checks; a risky build never goes live.
Written fast,
stalls at Kubernetes.
AI-assisted development increased the speed of writing code — not the speed of moving it safely to production. Once the code is ready, the real work begins.
The faster the team writes, the more scattered the infrastructure becomes.
Every team invents its own deploy method, its own security checks, its own DNS/TLS solution. As the number of applications grows, standards disappear and traceability declines.
Central control doesn't limit speed; it makes it last.
One platform, one security chain, one approval mechanism — it kicks in the moment a developer or AI agent says "publish." The standard stops being a constraint and becomes speed itself.
Say "go live";
let the platform handle the rest.
The developer, or the AI assistant working on their behalf, pushes the code; VibeHub runs the chain from Git to a scanned, approved, live Kubernetes service. Standard releases require no YAML.
The Git repository is authorized automatically; existing GitHub and Azure DevOps repositories can also be connected.
Container images are produced with ready-made build guides for 14+ languages/frameworks — React, Node.js, Python, Go, Java, .NET 10, Rust and more.
Secrets, static code analysis, dependencies/CVEs, image scanning and SBOM — before deploy; the pipeline stops on a policy violation.
The AI agent requests a deploy and human approval kicks in on first use — with scoped, time-limited and instantly revocable access (see 03 · AAT + HITL).
The service is published at an HTTPS address under the platform domain; the port is detected automatically, and no per-app DNS record is needed.
Three-stage health checks, live logs, automatic backups and approved restores — manage the live application from your browser, with role-based permissions.
Let agents work fast,
while you keep control of permissions.
Not a slogan, but a concrete mechanism: the Agent Access Token (AAT) + Human-in-the-Loop (HITL) authorization chain. Claude Code, Cursor and AiX all work through the same chain.
The token is valid only for the defined project — it cannot access other projects, cluster management, the registry or secrets.
Valid within a specific time window; automatically invalidated when it expires.
The agent requests access; the token is activated after an authorized person approves it in VibeHub.
Access can be revoked at any time, in a single step.
Every build, deploy, scaling event and terminal access is collected in a single audit trail.
Your assistant does more.
You always have the final say.
Claude Code, Cursor or AiX — your assistant connects to VibeHub by signing in, has you confirm the project before starting work, and manages only the settings you allow.
The assistant signs in to VibeHub; the access key is created automatically, no copy-pasting.
"I'll be working on the Support Portal project. Do you approve?"✓ User approved
Single project · closes at expiry · revocable anytime · every step in the audit trail.
Settings open to the assistant
The assistant manages email, AI access, app and health settings; the developer finishes the job from the editor.
The assistant cannot read secrets. Defining new addresses and identity providers is up to you; restarts and deploys happen with your approval.
The agent guide is refreshed with every release; the assistant works with the platform's current rules.
AI spend and live operations.
On one screen, per service.
Publishing isn't the end of the job. AI spend is visible per service and per model; the health, logs and backups of the live application are in the same console.
AI Gateway · monthly budgets sample data
Live operations sample data
Every build scanned,
every deploy traceable.
Security isn't a report, it's a deploy decision — the pipeline stops on a policy violation, and a risky build never goes live.
Secrets, static code analysis, dependencies/CVEs, container image scanning and SBOM — before deploy, with policy-based blocking.
Single sign-on with Entra ID and OIDC-compatible identity providers. Enterprise login via Auth Proxy without touching application code.
Role-based access control makes it clear who can do what at the tenant/project level.
Build, deploy, scaling, restart, terminal access, YAML changes and approved restores — all recorded.
The backup plan is ready once the component is installed; the target is your own S3 bucket, and restores require authorized approval.
Secrets are stored in Vault; AI assistants cannot read them.
Your application in your environment,
under your control.
VibeHub runs self-hosted: on your own Kubernetes cluster (BYOC) or in an isolated environment ATP operates for you — the architecture is the same in both models. It is installed with Helm and connects to your existing metrics system.
Your Kubernetes Environment
With BYOC, code, data and runtime stay on your own infrastructure; in the ATP-managed model, data residency is defined in the contract.
VibeHub Control Plane
Governance, automation and visibility layer. With BYOC, the control plane, policies and approval records all stay within your boundary.
Tenant, project, service —
one hierarchy.
Each tenant has its own namespace isolation, resource quota and identity provider; each project manages its own set of environments (dev/staging/prod).
platform/vibehubtenant/example-companyproject/support-portalsvc/support-portal-webRepresentative views
of the real interfaces.
The screens below are representative visuals that reflect the real layout and flow of the VibeHub interfaces — sample data is used.
Sign in with Enterprise SSO
Select your organization and sign in with single sign-on through your identity provider.
Service health and architecture
CPU/memory usage and namespace-based isolation for all services on one screen.
| Name | Status | CPU | Memory | Namespace |
|---|---|---|---|---|
| Sample Service A | Healthy | 57% | 12.4Gi/20Gi | tenant-a-svc |
| Sample Service B | Healthy | 26% | 1.1Gi/8Gi | tenant-b-svc |
| Sample Service C | Healthy | 29% | 2.9Gi/16Gi | tenant-c-svc |
| Sample Service D | Healthy | 56% | 3.6Gi/16Gi | tenant-d-svc |
Build & security chain
Every commit goes through the same chain; deploy stops on a policy violation.
Findings: 0
Findings: 0
Critical: 0
Critical: 0
Assistant connection
The assistant connects to VibeHub by signing in; the access key is created automatically, and no work starts without project confirmation. No need to copy instruction files.
session: VibeHub sign-in ✓ — access key automatic
confirm: "I'll be working on the Support Portal project. Do you approve?" → approved
settings: email, AI access, app (each enabled separately, off by default)
closed: secrets, other projects, cluster, registry
duration: single project · closes at expiry · revocable anytime
AI agent authorization & audit trail
Each token's scope, duration and status are visible; every action is recorded line by line in the audit trail.
| Project | Token | Status | Last Action |
|---|---|---|---|
| Sample Project A | vhaat_6e3e•••• | Locked | deployed · 9 days ago |
| Sample Project B | vhaat_e499•••• | Locked | build completed · 10 days ago |
| Sample Project C | vhaat_dc57•••• | Locked | security scan clean · 2 weeks ago |
VibeApps — internal discovery catalog
Discover the apps published in your organization: description, login link, owning team.
Who it's built for.
VibeHub meets the needs of four audiences with very different relationships to infrastructure, on the same platform.
AI-assisted developers with limited Kubernetes knowledge. The code is ready; the publishing know-how is missing.
Organizations with 5–50 developers. They have to keep a growing number of applications standardized and visible.
Teams working with many clients. They need per-client isolation and AI cost visibility.
Fintech, healthcare, finance. No deploy is valid without data sovereignty, an audit trail and an approval flow.
Not how the code is written,
but how it gets into production.
VibeHub is not a code generation tool — it manages how code (whether written by a human or an AI agent) gets into production securely and auditably.
If you have a static or edge-first, stateless frontend, VibeHub may be more than you need — VibeHub is designed for applications with stateful components such as databases/queues and multi-service architectures.
A catalog/portal layer doesn't replace runtime functions such as build/scanning/deploy/backup/agent authorization — those would still have to be set up separately.
VibeHub doesn't replace your cluster; it governs the standards by which applications produced by developers or agents enter that cluster.
VibeHub doesn't generate code — it takes every application you build (or an AI agent builds) to secure production, without lock-in to a framework.
A fixed platform license,
scaling by tenant and project scope.
The number of users or agents is not a billing item — it is only an adoption signal. Prices are for BYOC (self-hosted on your own infrastructure).
AAT + HITL governance model and a unified audit trail — AI assistants work under enterprise control from day one.
End-to-end, single-point traceability of the deploy chain; shortens the time to find the root cause of errors and incidents.
Significantly reduces the need for a separate platform engineering team to build and maintain an equivalent tool stack.
Publish applications,
not infrastructure tickets.
Let's publish your first real application together. With your own application, let's walk through the entire flow from Git repository to live environment.