The customer-facing features, improvements, and fixes making Rock8Cloud better every week.
Comment on an open pull request to request a preview environment for a service. The preview builds from the pull request's latest commit, and it tears itself down automatically when the pull request is closed.
A new blog post reflects on two months of building on our own platform, and a step-by-step guide shows how to connect Claude to the MCP server.
Custom domains are no longer removed when a service is scaled down for sleep and brought back up.
CPU and memory usage stay visible for past deployments. Charts now span the deployment timeline, with markers that jump to each deployment's history.
Admin and settings pages no longer overlap on mobile, the mobile sidebar opens reliably, and S3 credential rotation shows a proper progress animation.
Out-of-memory events now appear as markers on the deployment memory chart, and the warning clears on its own once the service recovers.
A metrics page and resource usage panel chart CPU and memory for each deployment over the last hour, 6 hours, or day, compared against configured limits.
Deploys no longer fail for projects with unconventional names, MCP custom domain queries return proper timestamps, and code review results refresh without a page reload.
Selecting a Dockerfile now shows a syntax-highlighted preview of its contents, including the detected exposed port.
Connected agents can update a service's repository, branch, Dockerfile, port, and health endpoint, and manage custom domains including verification.
Admins can grant or block individual AI models per organization, so teams only see the models they are allowed to use.
The dashboard header is usable on mobile, the add-service dialog keeps its buttons visible, and the landing page video plays on load.
Deploy a web-based PostgreSQL browser as an application service. Admin credentials and the database connection are generated automatically.
The new-deployment page uses a full-page layout with a footer that stays visible and clearer inline error messages.
Application deploy logs open the right tab, the billing page shows the correct subscription end date, expired invitations are listed separately, and application URLs resolve correctly.
Models are grouped by provider with icons and price tiers, and the agents tab jumps straight to the composer when there are no sessions yet.
Preview environments can be removed from the service menu, which tears them down. The next push to the branch recreates them.
Blueprints show a clickable screenshot of the deployed result before you clone the stack.
Link a service to a different GitHub repository from the build settings, without recreating the service.
The Strapi and Next.js blueprint now generates a shared preview secret, so Strapi previews render in the Next.js frontend with no manual setup.
Closing or merging a pull request closes its agent session cleanly. A follow-up prompt starts a fresh session that picks up from where the previous work left off.
Deployments finish in-flight requests before shutting down, so a rollout no longer drops traffic.
Pull request workflow settings are no longer lost when the workflow is saved.
A new article explores how agent-first primitives change the deployment workflow, with OG images updated across the site.
Blueprint descriptions no longer show stale data, cursor pointers work consistently across dropdowns and command items, and database version parameters and default S3 bucket sizes are corrected.
A dedicated landing page showcases the full MCP primitives catalog, helping you discover available tools and integrations.
The landing page header has been redesigned with consolidated platform pages and a cleaner menu structure.
Mobile navigation pages scroll properly, code-review prompts use the correct tools, and how-to/blog headings have consistent spacing.
Landing page copy has been updated and MCP tags on the platform page are correctly displayed.
Deployment pages now display inline MCP tag badges, making it easier to see which integrations are active.
Sidebar navigation uses a drawer pattern on mobile, and session details are now visible through a collapsible panel.
Subscription updates no longer race with concurrent writes, and the mobile sidebar navigation is more reliable.
The landing page has been redesigned with a new hero section, comparison section, pricing updates, and a platform band that highlights key capabilities.
Deploy a fully functional admin dashboard built with Elysia and TanStack, complete with managed PostgreSQL.
The service navigation rail, top bar, and build settings have been moved from a modal to a dedicated page with a clearer layout.
Custom agents can now stage and trigger other agent workflows in the same project, enabling multi-step automation chains.
Agent session transcripts identify whether each message was sent by you or the agent, making conversations easier to follow.
Blueprint connection URLs resolve correctly, agent session billing displays proper costs, and subscription cost translations are fixed.
Blueprint deployments with identical slot names now generate unique URLs automatically, preventing deployment conflicts.
Artifact files now show the uploader's name and link directly to the artifact for quick access.
Agent sessions show more descriptive completion messages.
Artifact uploads no longer time out when handling larger files.
Workflow trigger type can now be changed after the workflow is created, so you can switch a workflow from push to pull request without rebuilding it.
Manual deploy triggers run even during an active deployment, agent sessions handle unexpected model responses more gracefully, and the organization switcher tooltip shows only when text overflows.
The Strapi + Next.js blueprint now comes with step-by-step how-to guides and supports environment configuration for multi-environment setups.
Faster S3 public access toggling, reliable storage status, workflow description fixes, and better manual deploy redirects.
Define your own agents with custom system prompts and tool permissions, then run them as workflows triggered by pushes, pull requests, deployments, or build outcomes.
Each custom agent has a name, description, system prompt, and an allowlist of platform tools. Agent workflows keep triggers, service context, and result actions separate from the agent definition, so the same agent can commit a branch in one workflow and only notify in another.
Create a managed artifact bucket alongside your services, resize its storage from the dashboard, and connect it through environment variables.
Artifacts buckets get their own service view, deployment lifecycle, and configuration. Resize storage from the service overview and link credentials into dependent services like any other managed dependency.
Deploy a Strapi 5 CMS and Next.js 16 frontend together. PostgreSQL and S3 media are provisioned and wired automatically, with a published home page live on the first deploy.
The blueprint forks the monorepo, creates both services, provisions PostgreSQL and an S3 bucket, generates Strapi secrets, and sets environment variables to wire them together. No manual setup needed before the first deploy.
Connected agents can run custom agent workflows - define the agent, pick the trigger, and the agent runs when the event fires.
Agent sessions now show a diff sheet of changed files, so you can inspect what the agent modified before reviewing the pull request.
The diff sheet opens from the session transcript, showing added, modified, and deleted files with inline changes. Files changed by the agent are listed as a fallback when the full diff is still loading.
Agent sessions show a cleaner transcript, dedicated session pages, and a new session composer that lets you pick between coding and research before starting.
Monitor a repository service from its main environment, choose a health path, and see availability, response time, and outage history.
Use a platform URL or verified custom domain, then follow 30-day availability and daily incidents from the service dashboard.
Connected agents can inspect service availability, response times, and recent incidents without leaving the conversation.
Connected agents can create or update manual environment variables after confirming the exact values with you.
Coding and research sessions restore their workspace and conversation after suspension, so you can continue where you left off.
The repository state, conversation, and working context return together. Long-running tasks can pause without forcing you to brief a new agent from scratch.
Service pages now use a compact navigation rail, clearer environment context, route-based tabs, and dependable deep links.
Large changes are reviewed in focused groups, while code diffs load directly from the source for smaller, more accurate reports.
Reviews combine a plain-language summary, changed-file tree, architecture notes, guided diff walkthroughs, and findings pinned to exact lines.
Move from the high-level impact of a pull request into the exact code behind each finding without leaving the report.
Select review findings, add line-level instructions, and hand the batch to a coding agent on the pull request branch.
The agent works on the existing branch, pushes a fix, and triggers a fresh review. Follow-up instructions continue in the same session.
Build arguments work without matching Dockerfile declarations, temporary build interruptions retry automatically, and concurrent deploys no longer collide.
Push workflows can match branch patterns or tag patterns such as v* and release-*.
Start multi-turn coding or research sessions on a service, resume them later, and keep their branch and conversation context.
Choose a coding session for repository changes or a research session for investigation. Continue with follow-up prompts instead of starting a separate one-shot task.
A new prompt walks an agent through validating a repository, selecting a branch, creating a service, deploying it, and diagnosing failures.
The flow checks the Dockerfile and port before creating resources, then follows the deployment through completion and opens build logs when something fails.
Generate and deploy fresh S3 credentials from the service detail panel with a confirmation step for dependent services.
Mercury services now get more CPU and memory each at the same plan price.
Repeatedly crashing services can be scaled down automatically, with a dashboard warning and optional Slack, Discord, or webhook notification.
Agents can inspect and link environment variables, then provision PostgreSQL, Dragonfly, or S3-compatible storage in one step.
Credentials flow through linked variables without exposing their values. Agents can prepare several dependencies and redeploy the application once at the end.
Enable optional public read access for an object storage bucket from the dashboard or API.
Deploy complete starter stacks from a public catalog instead of assembling every service by hand.
Choose a blueprint, review its applications and dependencies, create the repositories in GitHub, and deploy the full stack into a new or existing project.
Project guides now cover Go, Java, Kotlin, Laravel, Next.js, PHP, and React starters.
Build output has simpler stage styling, clearer errors, complete timestamps, and stronger recovery when indexing is delayed.
Attach more than one custom domain to the same service.
Commit hashes in deployment logs now open the matching GitHub commit when available.
Set an HTTP path Rock8Cloud can use to determine when a repository service is ready and healthy.
The health endpoint travels with the service configuration across initial deployments, redeployments, and scaling changes.
Neptune services can run multiple application replicas for additional capacity and resilience.
Deploy the latest commit from a selected branch directly from the Deployments tab.
Configuration changes take effect without requiring a second manual deployment.
Export logs for offline investigation, incident notes, or sharing with your team.
Deploy repositories that do not yet include a Dockerfile by asking an agent to prepare one.
The agent detects the application stack, generates the required deployment files, validates the result, and returns the changes through the normal repository workflow.
Connect MCP clients such as Claude to Rock8Cloud and manage deployments through conversation.
Agents can discover organizations, projects, services, repositories, and branches, then create a service, deploy it, follow its status, and inspect code review results with your approval.
Organization and instance settings can enable or disable project scans, code reviews, coding agents, changelogs, chat coding, and landing-page design separately.
Pull requests show review progress and results in GitHub Checks, even when no notification channel is configured.
Create private S3-compatible storage alongside your applications and link its connection settings into dependent services.
Each bucket gets its own service view, deployment lifecycle, endpoint, and credentials, so object storage can be managed like the rest of a project.
Scale an idle service down and restore it later without deleting its configuration.
Sleep and wake actions apply across repository services and managed databases, while the dashboard keeps their current state visible.
Attach a domain you control to a deployed service and follow the DNS setup from the dashboard.
Rock8Cloud shows the required records, tracks domain configuration, and routes verified traffic to the selected service.
Pass build-time configuration into container builds from service settings.
Service names can include spaces and punctuation while stable internal URLs are generated automatically.
Connect Slack with OAuth and select exactly which channels each workflow should notify.
Billing shows your plan, current service usage, build minutes, and agent runs, with clear warnings before a limit blocks an action.
Compare available tiers, see which plan is active, and understand capacity before creating another service or starting agent work.
Create and edit push, pull request, deployment, analysis, review, and notification workflows from a dedicated service tab.
Choose the trigger, assemble the steps, and connect notification channels without writing pipeline configuration by hand.
Connect Slack from settings and use the workspace in project notification channels.
A redesigned dashboard brings project cards, recent activity, service health, and faster create-service actions into one view.
Search and inspect a history of deployments, services, projects, integrations, settings, and other organization actions.
Each record identifies the action, actor, affected resource, and time, giving teams a shared operational history from one place.
Ask an agent to implement a change in your repository, follow its progress, and receive the result as a pull request.
The agent works in an isolated branch, streams its progress to the dashboard, and leaves the final code ready for your normal review process.
Scan a service for architecture, quality, security, and maintainability insights, then review pull requests from the same dashboard.
Project analysis builds a broader understanding of the codebase. Pull request reviews turn that context into focused, line-level findings for each change.
Builds, deployments, database operations, analysis, and agent work now share a consistent status and log experience.
Choose a Dockerfile subdirectory as the build context for monorepos and nested applications.
Create another organization and move between organizations from the dashboard switcher.
Send deployment and workflow updates to Slack or a webhook endpoint.
Create a managed database alongside your services and link its connection settings into an application.
Rock8Cloud handles provisioning and connection details while linked variables keep applications configured as dependencies change.
Deploy pull request changes to isolated preview URLs before merging.
Each preview keeps its branch, deployment history, status, and live URL separate from the stable environment.
Service cards and details show whether a deployment is progressing, healthy, degraded, or failed.
Connect a GitHub organization, choose an accessible repository, and keep deployments updated through repository events.
Repository access stays scoped to the installation you select. Branches, commits, and deployment triggers stay connected without personal access tokens.
Build and rollout logs remain available after the live stream ends.
A dedicated onboarding state helps new organizations move from an empty dashboard to their first live service.
Rock8Cloud reads exposed ports from the selected Dockerfile and pre-fills the service port for you.
Invite teammates into an organization and let them accept through a dedicated invitation flow.
Organize related services into projects, create services inside them, and manage each deployment from a redesigned dashboard.
A project keeps applications and their dependencies together. Each service gets its own configuration, history, status, logs, and lifecycle actions.
Sign in with GitHub, choose a repository branch, and deploy without pasting repository credentials.
Build and service output are separated into focused tabs for easier troubleshooting.
Add application configuration and secrets during service setup and pass them into the deployed container.
Authenticated team members can see and manage deployments available to their access group.
Create an application together with a PostgreSQL database and its connection configuration.
Remove a deployed application and its generated configuration from the deployment list.
Deploy repositories that use a non-standard Dockerfile location, filename, or container port.
The proof of concept worked. A repository became a live service, marking the start of the Rock8Cloud story.
