-
Notifications
You must be signed in to change notification settings - Fork 0
FAQ
This page reflects the current NEXUS AI platform. For live plan limits, models, regions, and prices, use the dashboard and pricing page.
NEXUS AI is an MCP-native platform for building, previewing, deploying, and operating containerized applications. It combines an AI App Builder, source/image deployment, databases, storage, backups, secrets, logs, GitHub integration, REST APIs, a CLI, and an OAuth-protected MCP server.
No. Code generation is one entry point. Existing Git repositories, uploaded source, and container images can use the same deployment and operations platform without AI generation.
It provides a managed container runtime and also orchestrates supported AWS, Google Cloud, and Azure services. Enterprise/private installations may run in customer-controlled infrastructure.
Use nexusai.run/docs. This wiki focuses on stable concepts and public GitHub workflows.
Builder is a persistent conversation with preview, manual edits, attachments, click-to-edit, checkpoints, GitHub push, and repeated deployments. Deploy from Prompt is a one-pass generation-and-deploy workflow.
No. The preview is bundled in the browser and rendered in a sandboxed iframe. Infrastructure is provisioned only after an explicit deployment.
React single-page apps render directly. Full-stack Next.js + Prisma apps use mock data in preview and a real database when deployed. Unsupported packages produce a visible compatibility message instead of silently failing.
Yes. Images are sent as vision inputs after browser-side resizing. They can be used as design references or bug reports.
No. It sends a concise description of the selected element, visible text, classes/snippet, and ancestor path.
Yes. Each assistant turn and saved manual edit creates a full checkpoint. Restoring a checkpoint creates a new checkpoint without deleting history.
It can generate a Next.js/Prisma user model, password hashing, signed HTTP-only sessions, and route guards. Add authentication after the base app in a focused follow-up and review the generated security model before production.
Yes, through the NEXUS AI GitHub App. Repository creation depends on GitHub installation permissions and account type; pushing to an existing permitted repository is the most reliable path.
- NEXUS AI container runtime (
docker) - Google Cloud Run (
gcp_cloud_run) - AWS App Runner through the compatible CLI identifier
aws_ecs_fargate - Azure Container Apps (
azure_container_apps)
Availability varies by plan and provider configuration.
The NEXUS AI docker runtime supports the complete Compose-style stack. Cloud Run, App Runner, and Container Apps are single-container targets; use supported standalone managed databases there.
Yes. NEXUS AI detects common frameworks and can build source repositories. Supply custom install/build/start commands or a Dockerfile when detection is not enough.
Read PORT when available, listen on 0.0.0.0, and ensure the declared service port matches. Listening only on localhost is a common cause of an unhealthy deployment.
Yes on the NEXUS AI container runtime. Use --worker-command and optionally --worker-name in a source deployment.
Stop is a soft lifecycle action. Start rehydrates the same deployment configuration without a source rebuild. Delete is the destructive deployment action.
Use the dashboard, nexus deploy scale, REST API, or nexusai_deploy_scale. Provider and plan limits apply. A single-writer filesystem volume is not a distributed filesystem; use a bucket or database for shared replica state.
Yes. Set --auto-destroy during source deployment or use nexus deploy auto-destroy --in, --at, or --off later.
Yes. Add the domain, publish the exact returned DNS records, verify ownership, and wait for certificate issuance. Proxy services can interfere with ownership or TLS validation.
PostgreSQL, MySQL, MongoDB, and Redis are supported as docker deployment services.
NEXUS AI supports local standalone databases and configured AWS RDS, Google Cloud SQL, and Azure database services. Exact engines, versions, classes, and regions depend on provider availability.
Standalone managed databases can be attached according to platform/provider rules. Deployment sidecars belong to one deployment stack. Use explicit environment prefixes if multiple database connections are attached.
A volume is a mounted filesystem path for one deployment. A bucket is S3-compatible object storage that can attach to multiple deployments. Volumes suit SQLite/filesystem state; buckets suit uploads and shared objects.
Attachments change desired configuration. Redeploy the app after attach, detach, or credential rotation.
Deployment services support engine-appropriate backups, download, in-place restore, cross-service restore, and scheduled backups. Standalone databases use snapshots and restore into a new database.
No. Code releases and stateful resources have separate histories. Take a database backup before risky migrations and restore state explicitly.
Yes. Register it as an external PostgreSQL source. Use the Supabase Session Pooler on port 5432, the pooler username, and SSL mode require.
curl -fsSL https://nexusai.run/install-mac.sh | bash # macOS
curl -fsSL https://nexusai.run/install.sh | bash # Linux
npm install -g nexusapp-cli@latest # Node.js 18+Create a scoped access token and set NEXUSAI_TOKEN. Use NEXUSAI_API_URL only when targeting a non-default installation.
Yes. Connect https://mcp.nexusai.run/mcp with OAuth. The current server exposes 74 tools across Builder, deployments, secrets, domains, databases, backups, storage, support, and usage.
Yes, under io.github.nexusrun/nexus-ai.
Any compatible HTTP MCP client can connect. Documented setups include ChatGPT, Claude Desktop, Claude Code, Cursor, and Codex.
Scopes separate read, mutation, and deletion privileges, and destructive operations are placed behind confirmation-aware tool categories. Review every resolved resource ID before approving an operation.
Yes. User workflows generally use a JWT; automation uses scoped access tokens or OAuth. See MCP and REST API.
Secrets and provider credentials are encrypted at rest. Deployment secrets are injected at runtime rather than committed into source or images.
Normal list operations return metadata, not plaintext values. Resource-specific credential reveal operations, such as bucket credentials, require elevated permission and are audit logged.
No. Redeploy affected workloads so new process environments receive the value.
The active role model includes Owner, Admin, Member/Developer, Deployment Manager, Auditor, and Billing Manager responsibilities, with resource actions further constrained by scopes and provider rules. Use the current permissions matrix in the docs.
The platform provides HIPAA-aligned controls and healthcare/enterprise offerings. Whether a workload is compliant depends on configuration, contracts, operational procedures, and the services used. Review the HIPAA page and obtain the necessary BAA before processing PHI.
Report suspected vulnerabilities privately through customer support. Do not disclose sensitive vulnerability details in a public issue.
It provides compatible model endpoints plus provider routing, fallback, caching, budgets, rate limits, analytics, credential isolation, guardrails, and audit records.
Usually not. Change the SDK base URL and gateway key. OpenAI-compatible traffic uses /v1; Anthropic-compatible messages use /v1/messages.
No. AI Gateway routes model inference traffic. MCP lets an agent operate the NEXUS AI deployment platform.
Plan names, allowances, and included providers can change. Use nexusai.run/pricing for current information. The dashboard displays the active organization plan and usage.
Customer-owned AWS, Google Cloud, or Azure resources may be billed directly by that provider. Review both NEXUS AI and provider billing before enabling production workloads.
- Customer support
- In-app support tickets
- MCP support-ticket tools for authorized agents
- Platform status
Inspect full build logs, lockfile consistency, runtime version, private registry credentials, native build dependencies, and transient registry errors. Keep the build context small.
Check 0.0.0.0 binding, port, start command, required secrets, migration duration, dependency connectivity, and runtime logs.
Confirm the connected service identity has the documented provider permissions and that IAM propagation has completed. Inspect the provider build/runtime logs as well as NEXUS AI logs.
Check organization role, access-token/OAuth scope, plan availability, provider configuration, and whether the resource belongs to the active organization.
Use the Knowledge Base and the troubleshooting sections in the docs.