Nova works on the stack you already run, with every production action scoped, approved, audited, and reversible.
Everyone diagnoses. The control plane for safe action is the moat.
Currently onboarding early design partners.
Your team could be here. Become a design partner.
Connectors and a model can guess a cause. Diagnosis is becoming a commodity. The hard half is earning permission to act in production.
A model with connectors can describe what broke. Letting it change production needs a scope it cannot exceed, a blast radius ceiling, and a rollback. Nova is built around the control layer.
Approval is a runtime primitive, not a button. Every action carries identity, scope, policy, blast radius, rollback state, and expiration, written to an append-only audit trail.
A chat session forgets when it closes. Nova accumulates operational knowledge in the platform: what worked, what was ruled out, and your topology, so each incident starts warmer than the last.
Connectors call APIs and hope. With Astro Platform, Nova enforces what is safe to run at the runtime itself, with a first-party map of your environment. Nova starts on your existing tools. Astro Platform unlocks governed execution.
The model was never the bottleneck. The control layer was. Everyone can diagnose. The control plane for safe action is the moat.
The value is the loop: investigate, plan, approve, execute, observe. It stays stable as skills and connected systems expand.
Nova forms hypotheses, collects evidence, identifies findings, and proposes remediation paths.
Requests become multi-step workflows with dependencies, approval points, and visible execution paths.
Sensitive write workflows use approval gates, then stream progress through the same operator model.
Nova acts through skills: Astro Platform, GitHub, cloud APIs, and CLI workflows. It connects to your systems instead of pretending to be them.
Managed Nova with zero setup. Browser, CLI (astroctl nova), and Slack are all interfaces to the same managed backend. Start in the browser, continue from your terminal, get notified in Slack. Sessions and context carry across.
Self-hosted via Docker Compose. Use local model execution when your operating model requires data to stay inside your boundary. Prompt-to-production deployments, browser UI, and MCP Marketplace support.
Your AI platform engineer, in your editor. Investigate incidents, manage clusters, and debug apps directly from Claude Code, Cursor, or VS Code. Hosted MCP endpoint: just add the config and go.
Real recordings of the real product. Investigate, propose, and act under governed approval.
Nova applies the fix when it can, and holds off when it should — honest about every write.
Nova reads freely. Nova writes only through the runtime.
Ask Nova anything about your infrastructure
Nova becomes useful fast when it can collaborate with your team, act through skills, and reason with real platform context.
Use Nova for approvals, notifications, async follow-up, and team visibility through the Slack skill.
Astro Platform is the runtime where Nova goes deepest: clusters, applications, rollouts, audit, and governance.
Nova uses skills for GitHub, cloud cost, infrastructure actions, and external systems instead of hardcoded integrations.
Nova acts through governed skills. Connect the platform skills your team already uses in one click, or bring any MCP server. However a skill is added, Nova treats it the same: scoped, policy checked, and audited.
AWS, Google Cloud, Azure, Kubernetes, Helm, Terraform, GitHub, Slack, Datadog and more, plus the Astropulse Platform itself. Authenticate once and Nova can use them, with every action scoped, policy checked, and written to an audit trail.
Paste a Model Context Protocol server URL over Streamable HTTP, add a bearer token or API key, and Nova discovers its tools in seconds. Once added, your server is a first-class skill, indistinguishable from the ones we ship.
Nova investigates like an SRE: forms hypotheses, correlates logs, metrics, traces, and recent deploys, and ranks the likely root cause, getting sharper with every incident. Unlike read-only tools, it can then act on the finding under the control plane.
AI expresses intent. The runtime decides what runs. Every action runs inside a scope it cannot exceed, with human approval for sensitive work. Execution is streamed live and written to an append-only, per-org audit trail.
astroctl nova launches a full interactive chat in your terminal. Paste logs, ask questions, investigate incidents, and resume prior sessions. Slash commands and keyboard shortcuts make it a native operator workflow.
$ astroctl nova$ astroctl nova --continue$ astroctl nova --resume$ astroctl nova sessions list --limit 25Nova works with your existing tools out of the box. Astro Platform is the runtime where Nova gets built-in cluster, application, rollout, cost, audit, and governance depth, and where safety is enforced at the runtime itself. It unlocks governed production execution.
Today, Nova already ships the operator model: investigations, plans, approvals, and execution, with deeper control through the Astro Platform runtime and the CLI.
The long-term story is broader platform engineering coverage, but the user does not need to relearn the product every time a new skill or integration appears.
Nova is the AI platform engineer. Skills connect it to your systems. Astro Platform is the runtime that unlocks governed production execution.