Container updates and private previews on your infrastructure.
Prontera manages Docker update workflows on your servers. We’re building toward private, full-stack previews that your team—or a coding agent—can request through an optional Cloud control plane.
The public Apache-2.0 Core release is planned. Prontera Cloud is not yet available.
Planned Cloud workflowFIG. 01
A changeDeveloper or agent
Verified artifactsBuilt by trusted CI
Approved preview template
Your preview VPSCustomer-owned
One private application. The whole stack.
FrontendAPIPreview database
Isolated preview dataBounded lifetime
Private reviewInvited reviewers only
Expiry & cleanupSeparate approval → Release Later
Concept illustration, not a live environment
A preview of the workflow—not a live environment.
Customer-controlled infrastructureExplicit policies and approvalsAn independently useful Core
01 /Two parts. Clear responsibilities.
Start locally. Add hosted coordination when you need it.
Prontera Core and Prontera Cloud have different responsibilities. The Core performs local operations. Cloud is planned to coordinate projects, reviewers, and connected hosts without becoming the place your application runs.
Development
Prontera Core
Self-hosted · Development build
Watch container images, apply update policies, check health, and inspect the outcome from a local interface or CLI. The intended open-source release remains useful without a Cloud account.
Connect existing servers, request private previews, give reviewers access, and let coding agents request permitted operations through the same control plane used by your team.
The Core brings image checks, update policy, replacement, health validation, and recovery into one workflow.
Development
01Watch
02Evaluate policy
03Replace
04Check health
05Inspect / recover
Watch image changes
Check OCI image identities on a schedule or when an operator requests it.
Decide when changes run
Use notification-only operation, approval modes, maintenance windows, and hooks instead of treating every new image as permission to deploy.
Check the result
Evaluate the configured health gate after replacing an existing watched container.
Recover when an update fails
Attempt to restore the previous captured image and runtime configuration when a health-gated update fails.
Review configuration changes
Validate, compare, apply, and export prontera.yaml configuration.
Inspect operations locally
Use the CLI or embedded web interface to inspect watches, update history, and operational state through the daemon API.
These capabilities exist in the development code. Production qualification, private-registry pull authentication, and interrupted-update recovery are still being completed.
A recovery boundary: restoring a container image does not undo database migrations or application-data changes.
Planned for Prontera Cloud
Review the running application, not only the diff.
Keep your coding tools and CI. Prontera’s planned preview workflow starts with their output and gives a reviewer a private application to open.
Your CI builds the application images. Prontera verifies the repository revision, workflow, and immutable artifact identities before accepting the preview request.
An approved template defines the application, supporting services, resource limits, and lifetime. A designated customer preview node runs the stack with isolated preview data—not production database contents.
The reviewer signs in through Prontera Cloud and opens the environment on a dedicated Prontera preview domain. Access is granted to the specific environment rather than inherited from a public link.
The boundary at each step
Change
Eligible sources come first.
The initial native-preview model requires eligible customer-controlled revisions. Untrusted execution needs a later, separately qualified isolation path.
CI artifacts
Build outside the target host.
Your existing CI builds the application images. A build alone does not establish trust: the repository, workflow, revision, and immutable artifact identities need verification.
Private preview
A designated place for temporary workloads.
Approved templates define the stack and its limits. Preview data stays separate from production data, and the application runs on designated customer infrastructure.
Review
Review access has a specific scope.
Cloud-backed reviewer identity, preview access, and the application’s own login have separate responsibilities. Sharing a review does not grant infrastructure credentials.
After review
Ordinary lifecycle
Clean up the preview
The preview expires or is removed when its pull request closes. Prontera tracks the resources it created and reports incomplete cleanup instead of silently leaving them behind.
Later Optional approved release
Decide what happens next
Review the change, request another revision, or follow a separately approved release workflow when that capability is available. A working preview does not automatically authorize a production deployment.
The preview lifecycle does not depend on promoting a release. Cleanup is required either way.
Cloud capabilities below are planned.
One place to see the change, the environment, and the decision.
Preview the complete application
Bring up the frontend, API, and supporting services together using reviewed templates and synthetic development data. A successful preview should be usable for review, not simply a container marked as running.
Give reviewers specific access
Use Cloud-backed identity to authorize a reviewer for the right project and environment. Keep the preview session separate from infrastructure credentials and from the application’s own login.
Keep temporary infrastructure temporary
Apply preview lifetimes and resource limits. Track cleanup across containers, networks, volumes, and routes, including when a node disconnects.
Let an agent request an operation, not a root shell.
An agent should be able to ask for a preview, inspect a failed operation, or request a release plan without receiving unrestricted control over the host.
Prontera’s dashboard, API, CLI, and MCP interface are intended to use the same permission and operation model. Permissions will specify the projects, environments, and actions an agent may use. Production approval remains a separate decision.
No coding agent is required. The same workflow is intended for people and CI.
Prontera CloudPlanned
Human dashboard
CLI / API
MCP
One permission and operation model
Projects · Environments · Actions
Scoped operations
Prontera Core
Local execution boundary
Planned shared authorization. Production approval remains a separate decision.
Example request
“Request approval for this release plan.”
Illustrative request, not an implemented MCP tool.
05 /Integrations, with their status attached
Use the tools you already work with.
Prontera starts with Docker and existing image workflows. Additional integrations follow the same rule: support must be explicit, testable, and accurately described.
Runtimes and installation
Docker Engine on LinuxDevelopment
Present in the development Core. Supported production combinations will be published after qualification.
Podman: rootful, followed by rootlessLater
Later, separately qualified runtime tracks. Docker compatibility does not establish Podman support.
Image discovery
OCI registries and ECRDevelopment
OCI image discovery and ECR-specific integration code are present in development. Private-registry engine authentication still requires remediation and qualification.
Source and CI
GitHub and trusted CI artifact handoffPlanned
Planned first Cloud workflow for eligible repositories and verified CI artifacts. Other CI systems are not automatically promised.
Private ingress
Cloudflare TunnelPlanned
Selected for the planned managed preview path. This is not an existing partnership or a finished integration.
Self-managed notifications
Slack, Discord, generic webhooks, and SMTP exist in the development Core. Delivery hardening is ongoing.
Development
06 /The execution boundary
Your workloads stay on your infrastructure.
The local Core is the execution boundary. In the reference architecture, a constrained proxy—not the Prontera daemon—holds the Docker socket.
The Cloud design adds an outbound node connection, scoped operations, and reviewer access without requiring a publicly exposed Docker API.
Reference architectureConcept diagram
Prontera Cloud
Planned
Identity, scoped operations, and reviewer access.
Optional hosted coordination
Outbound node connection
Planned · No publicly exposed Docker API
Customer-owned infrastructure
Local Core execution boundary
Development
Prontera daemonLocal operations
Constrained proxyHolds the socket
Docker socketDocker Engine
Watched containersCustomer workloads
Native updates manage existing watched containers. Production qualification remains in progress.
Isolated preview data. No production database contents.
Trusted CI artifactsOutside the target host
Eligible customer-controlled revisions and verified immutable artifacts are required for the planned native-preview path.
The constrained proxy—not the Prontera daemon—holds the Docker socket in the reference architecture. Preview and production authority remain separate; a working preview does not authorize a production deployment.
Add identity, node enrollment, read-only visibility, and private application previews.
Planned
03
Agent operations and releases
Add scoped MCP actions and separately approved release workflows after their execution and recovery boundaries are qualified.
Planned
04
Broader compatibility
Qualify Podman and selected external execution providers against real customer needs.
Later
08 /A simple business model
A useful self-hosted Core. Optional paid Cloud coordination.
The planned public Core release will use Apache-2.0 and operate without a Cloud account. Cloud plans will cover hosted coordination, collaboration, and operational convenience—not the customer’s server bill.
We are preparing a small design-partner pilot before an invitation-only beta. Prices and usage allowances are still being evaluated.
Public Core release planned · Cloud is not on sale
09 /An independent project
Built around an independently useful Core.
Prontera starts with a practical problem: keeping existing Docker applications updated while making changes easier to inspect and recover from.
The planned Apache-2.0 release is intended to remain useful on its own. Prontera Cloud adds hosted coordination for people who need a shared workflow across projects, reviewers, and hosts.
Prontera is being built by Ramon Oliveira. The project is in early development, and the code repository is currently private while release preparation continues.
Not yet. The Core has working development functionality, but production qualification and important recovery and authentication work remain. Cloud access is not available yet.
Can I use it without Prontera Cloud?
The intended public Core release is independently useful and self-hosted. Cloud is optional hosted coordination, not a required account for local operation.
Will Prontera host my application?
The planned native workflow runs applications on your designated servers. Prontera Cloud coordinates operations and access; it is not a replacement hosting provider.
Do I need an AI agent?
No. Agent access is an additional interface to the planned workflow. Developers and CI can use the same operations without an AI model.
Will agents receive SSH or unrestricted Docker access?
That is not the product model. Planned agent access is limited to named operations and resource scopes, with separate approval where required.
Can anyone’s pull request run on my server?
Not in the initial native-preview model. The plan requires eligible customer-controlled revisions, trusted CI evidence, and approved templates. Untrusted execution needs a later, separately qualified isolation path.
Does rollback restore the database too?
No. Replacing an application image does not undo migrations or data changes. Application-data compatibility and recovery need their own plan.
Is Podman supported?
Not yet as a qualified runtime. Docker is the first target. Rootful Podman, then rootless Podman, are later qualification tracks.
What happens to a preview after review?
The design includes removal when its pull request closes or its lifetime expires. Cleanup must report incomplete work instead of implying that every resource was deleted.
Are you building a sandbox provider?
No. External sandbox or microVM services may be integrated later. Prontera is not building a competing microVM deployment platform.
10 /Join us at the beginning
Help shape the first preview workflow.
We are looking for developers and small teams already running containerized applications on their own servers. Join the waitlist to hear when the initial pilot and closed beta are ready.
A small design-partner pilot. Then an invitation-only beta.