BROCS
In Run Overview
R / SRE, PaaS, hybrid cloud

Run

The runtime underneath AI work: where apps and agents execute, the data and secrets they need, and the portability to run all of it where the data already is.

View as Markdown
Runtime where apps and agents actually execute
Data and state provisioned, owned, backed up
Secrets injected at runtime, never in code or chat
Provisioning self-service, provision-build-deploy as one motion
Portability the runtime goes where the data goes

Run covers the execution environment for AI applications and agents. It includes network reach, data stores, session credentials, provisioning, and placement.

A Build product that leaves deployment to the customer also leaves the runtime and its operating cost with that customer.

Runtime requirements

Long-running agents are stateful processes that may hold session credentials, execute generated code, and read untrusted content. Stateless service platforms may need a separate execution pattern for that workload.

Placement also matters when policy, contracts, or data volume prevent data movement. The same runtime may need to operate in several approved environments.

Operating tests

  1. Identify the current execution environment for each long-running agent, including its network and assigned permissions for cloud services and databases.
  2. List every store written by AI applications, including vector indexes and prompt logs.
  3. Test a second placement and record the dependencies that prevent it from running.

Tradeoffs

Isolation adds work to the Build path and provides necessary Secure boundaries. The Replit incident involved an agent with production credentials, missing environment separation, and a code freeze that the runtime did not enforce.

In this capability

  • Runtime: Execution environments for AI applications and agents, including containers, functions, jobs, and long-running sessions.
  • Data and state: Databases, object storage, vector indexes, and session state for AI workloads: provisioned on demand, owned, and backed up.
  • Secrets: Runtime secret injection for AI workloads, with prompt and trace redaction and short-lived session credentials.
  • Provisioning: Self-service infrastructure for AI work, from provisioning through build and deploy in one workflow.
  • Portability: Running the same workload across hosted and customer-controlled environments when data placement requires it.
R -> O Next capability: Observe. Measure system behavior and quality.