Skip to main content
Every Superblocks instance runs the same platform. Your deployment model decides where it runs:
  • Dedicated Cloud: Your single-tenant instance runs in a dedicated environment that Superblocks hosts and operates on AWS.
  • Managed BYOC: Your single-tenant instance runs in your AWS account.
In both models, a few shared services run centrally in Superblocks Cloud and serve every instance.
For security reviewers: This page summarizes where each part of the platform runs, what data it stores, and what leaves your instance. To request the SOC 2 report and other security documentation, visit the Superblocks Trust Center.

Where each part runs

Each model’s page has a diagram of the full layout: Dedicated Cloud and Managed BYOC.

Platform services

The Superblocks platform is made up of these services. Each one runs inside your instance, so it’s single-tenant in both deployment models.

Workload isolation

  • DevBox and RunBox: Each sandbox is a single-use pod, isolated with gVisor.
  • AI agents: Agents, including the Build Agent, run in microVMs on Amazon Bedrock AgentCore, in the control plane.

Control plane

The control plane is where IT and Security govern your instance. From it, they control integrations, infrastructure policies, and identity and access, and monitor everything in production.
  • Platform services: The Identity and Access Gateway, DevBox, the Skills Registry, CodeGuard, the control plane side of the Infrastructure Policy Engine, and Watchtower.
  • Control plane server: The entry point and system of record for your instance. It manages app code, metadata, configuration, and DevBox sessions, and coordinates platform services such as the Skills Registry and CodeGuard.
  • Control plane database: Stores platform data, including AI prompt and response history.
  • Build Agent: Builds apps by running coding agents such as Claude Code, Codex, and OpenCode. It runs in a microVM on Amazon Bedrock AgentCore and works on the app’s code in DevBox. It sends model requests to the Superblocks Inference Gateway.
The Identity and Access Gateway works with the global identity gateway, a shared service that routes sign-in to your SSO provider.

Alongside the control plane

Data planes

Data planes run your apps’ backend code and connect to your data sources. Each instance has separate non-production and production data planes, and you can add more, such as separate development, staging, and production data planes. You can also add data planes in other regions, closer to your data. See Data planes for placement and connectivity. Each data plane includes:
  • Platform services: The data plane side of the Infrastructure Policy Engine, RunBox, and the Integration Gateway.
  • Managed database: App databases, on Amazon Aurora.
  • File storage: App file storage, on Amazon S3.

How requests flow

Each flow lists the hosts it talks to, where its data is stored, and whether any of it leaves your instance. In Dedicated Cloud, your instance runs in a dedicated environment that Superblocks hosts and operates on AWS. In Managed BYOC, it runs in your AWS account.

Sign-in

  1. A user opens your instance URL, such as https://{tenant}.superblocks.com.
  2. The global identity gateway at login.superblocks.com routes authentication to your SSO provider. It stores the user’s identity information.
  3. The user returns to your instance with a verified identity, and the Identity and Access Gateway applies the permissions your admins assign.

Building apps

  1. A builder works in the editor, which runs in DevBox: a single-use, isolated sandbox in your instance that holds the app’s code and dev server.
  2. The Build Agent runs in its own microVM and writes and tests code in DevBox. It sends model requests to the Superblocks Inference Gateway, where inference providers don’t retain prompts or responses or train on your data.
  3. The Build Agent can search the web, for example for library and API documentation. In Managed BYOC, AWS Network Firewall can limit which sites it reaches.
  4. Package installs go to your npm registry, either public npm or your private registry, through the Superblocks npm proxy.
  5. To explore data schemas and sample data, the Build Agent uses tools that reach your non-production data sources through a non-production data plane. Your team decides which data it can reach. See Shared responsibility.
App code and AI prompt and response history are stored in the control plane, and development databases and files in non-production data planes. Only model requests, package installs, web searches, and lookups to your own data sources leave your instance.

Deployed apps

Every request to a deployed app is checked twice: at the CDN before the app’s frontend loads, and at the data plane before its backend code runs.
  1. When a signed-in user opens a deployed app, an edge function on the CDN checks that they’re authorized to access that app before the frontend loads.
  2. The app’s API calls go to the production data plane, which checks the user’s access on every request before RunBox runs the backend code.
  3. The Integration Gateway applies your access policies, so each user and app reaches only the business systems it’s allowed to.
App databases and files are stored in the production data plane, on Amazon Aurora and Amazon S3. Data leaves your instance only to reach the data sources and SaaS services your apps connect to.

Upgrades

Superblocks plans, executes, and supports every upgrade. Your instance pulls Superblocks container images from the Superblocks registry, and the deployer applies the upgrade from inside your instance. No app data leaves your instance during an upgrade. In Managed BYOC, the deployer runs in your AWS account, so the management plane never changes your account directly.

Telemetry

Your instance produces two kinds of telemetry: app telemetry and platform telemetry. They come from different sources and are handled differently. Only platform telemetry is sent to Superblocks.

App telemetry

App telemetry comes from your apps and agents: the logs, metrics, traces, and audit logs of the code your team writes, in DevBox while building and in RunBox once it’s published. Telemetry Pipelines enforce your telemetry requirements and route app telemetry to the destinations you approve, such as your data warehouse, SIEM, and monitoring tools. App telemetry stays in your instance and your approved destinations. It isn’t sent to Superblocks.

Platform telemetry

Platform telemetry comes from the services that run the platform, not from your apps. It’s handled separately from Telemetry Pipelines. Before platform telemetry leaves your instance, personal and sensitive information is removed. Superblocks receives it in its central monitoring and analytics service, and uses it to monitor platform health, validate upgrades, and troubleshoot with your team during support cases.

Telemetry in Managed BYOC

In Managed BYOC, telemetry is also written to CloudWatch and AWS X-Ray in your account, and you can send logs, metrics, and traces to your own observability destinations. Adding a destination doesn’t change what Superblocks receives.

Encryption

In both models, traffic is encrypted in transit with TLS 1.2 or higher, and data is encrypted at rest.