NEXUS AI

What it is

NEXUS AI describes itself as "an agentic AI app builder and full-stack deployment platform." You describe a product, the builder writes the interface and application code, checks it in a live sandbox, connects managed services and deploys it. It is built for founders and developers who want a prototype to become a running service, for developers who already work in Cursor, Claude, Codex or GitHub, and for teams that want AI agents to operate the platform through its CLI, REST API and MCP tools. The pages do not say where the company is based or when it started.

What you build and how

The core is the NEXUS AI App Builder. The homepage says it "creates and edits project files, runs checks in an isolated development sandbox, inspects errors, and revises its work before deployment." In the Builder you choose an enabled provider, describe the app, and review Preview and Code tabs. You refine by chat, file edits, screenshots or click-to-edit, and the Builder keeps checkpoints. A sample prompt on the deploy page asks for a support portal with ticket submission, status tracking and an admin view using Next.js and Prisma, with loading, empty and error states, and authentication added in a follow-up.

There are two kinds of preview. The instant browser preview is for iterating on the interface. A real full-stack development sandbox covers routes, forms, packages and database-backed behavior. The page is explicit that the instant preview does not run a production database, so you test server behavior in the sandbox and verify real data after deployment. You can also publish a frozen preview, or remix a shared app into a new project, and you can import an existing project from GitHub or bring in a local one.

What gets deployed is more than a web app. A deployment can include "the app container, PostgreSQL, Redis, MySQL, MongoDB, background workers, persistent volumes, S3-compatible buckets, logs, health checks, custom domains, backups, scaling, and rollback." The company's typical example is a web app, a PostgreSQL database, a Redis queue and a background worker running as one service group on a shared network, so the app can reach the database by an internal hostname such as postgresql or redis. Storage includes persistent volumes for apps that write to a path such as /data, plus S3-compatible buckets for uploads. Database backups can be restored in place or into another compatible service in the same organization. The homepage also names Stripe among the managed pieces.

The deploy page gets practical about why AI-written apps fail in production, and how the platform handles each case. An app that listens on localhost builds but never answers, so the process must listen on 0.0.0.0 and the configured port. A husky or prepare script can break npm install, so generated Node.js Dockerfiles set HUSKY=0 and CI=true. With no Dockerfile, NEXUS AI generates one or uses the Dockerfile at the repository root. If a Builder deploy fails, one click sends the build error back to the agent to fix and redeploy.

Code comes in and goes out several ways. You can push files to a Git repository, download a ZIP or use the Builder directly. Assistants such as ChatGPT or Claude chat can hand over up to 20 files into the App Builder. From the command line, a source deploy looks like nexus deploy with flags for services and a worker command, then nexus deploy status and nexus deploy logs --follow. The page says the documented example builds from a Git repository, while the dashboard's Deploy Full App option can also take a local folder or ZIP.

For AI agents, the company says it exposes 74 MCP tools. An MCP-compatible client connects to its MCP server and approves OAuth scopes. The agent can push complete files into the Builder with nexusai_builder_push for review, which saves a checkpoint and returns a preview URL but does not deploy, or deploy a Git repository with nexusai_deploy_source, then read nexusai_deploy_status and nexusai_deploy_logs.

NEXUS AI also runs its own products on the platform: NEMT OS, an operations platform for Arizona AHCCCS non-emergency medical transport providers covering trip scheduling, dispatch, live GPS and claims; ZOLLO, an AI campaign platform that turns a brief into an email sequence, social posts, an audience plan and a schedule; and Closr, a commission-based sales marketplace with Stripe-handled attribution and payouts. The company says these are first-party products, not third-party customer deployments.

Evidence from the pages

The deploy page publishes a plan table. Free costs $0 and includes 1 deployment, for testing, plus the App Builder with 500K AI tokens a month. Starter is $29 a month with 2 deployments, deploy targets of NEXUS AI or your Google Cloud account, a custom domain, 5M AI tokens and 3 team members. Pro is $149 a month with 5 deployments, and adds AWS, Google Cloud and Azure targets. The page also says Pro can deploy into your own AWS, Google Cloud or Azure account using Cloud Run, App Runner or Container Apps in the region you choose, and that rollback is available on Pro and higher.

For a platform that runs entirely inside your own network, the page says NEXUS AI Enterprise deploys in your cloud account or on-premises. The homepage says the platform includes tenant isolation, role-based access, encrypted secrets and backups, audit logs, and deployment options for AWS, Google Cloud and Azure. It does not name a compliance standard such as SOC 2, and it does not publish customer counts. The homepage links to Enterprise and Pricing pages, but no enterprise price appears.

Support is described by what exists: documentation, a knowledge base, customer support and a contact form. The pages state no response times.

Who it suits and who it does not

NEXUS AI suits a developer or founder who wants the whole path covered, from generated code to a database, a worker and a live URL, without stitching together separate cloud consoles. It suits people who already build with Cursor, Claude or Codex and want a place to deploy that code, and teams that want agents to build, deploy, scale and back up apps through MCP.

It suits non-developers less well. The workflow includes Git, a CLI, Dockerfiles, ports and logs, and the page's own advice is to review files before you deploy. Real limits sit in the plans: the Free plan allows 1 deployment for testing, and cloud targets of your own begin on Starter and Pro. The instant preview is not a production database, so full-stack behavior has to be tested separately. And the pages do not describe a drag-and-drop editor, so a team that wants visual app building without touching code should look at other categories.