Glossary
This glossary lists terms that other pages use without a full definition. Each entry has one line and a link to the page that defines the term. For a manifest kind, see the Manifest Reference. For the manifest, the three axes and the state ledger, see Core Concepts.
| Term | Meaning | See |
|---|---|---|
accessControl | The block that grants an application or an agent access to named resources. | Security Framework |
| Access tiers | The administrators, contributors and viewers levels of human access. | Access Tiers |
| ADK | Google's Agent Development Kit. Beck and the agents that you declare use it. | Agentic Governance |
| Binary Authorization | A policy that lets a project run only images that have a signed attestation. | CI/CD |
| Blast radius | The set of resources that a change can break. | How Beck Works |
CODEOWNERS | A GitHub file that names the owners who review a path. The platform writes it from your manifests. | Security Framework |
computed | The part of a manifest that the engine writes. Never edit it. | Core Concepts |
| Conventional Commits | A message format (feat:, fix:) that sets the version bump. | CI/CD |
| Core Project | The Google Cloud project that holds shared services. | Core Concepts |
| Drift | A difference between a resource and its manifest. The engine removes it on the next run. | Access Tiers |
| Egress allow-list | The allowedEgress list of a project. It names the outside hosts that apps can call. | Configure Egress Rules |
| Engine | The Go program that reads manifests and calls the Google Cloud APIs. | Architectural Overview |
| Hibernation | A schedule that removes compute from a non-production project. | Control Costs |
| JIT graph | The in-memory dependency graph that the engine builds on each run. | Architectural Overview |
| Managed identity | The scoped identity and tokens that the engine gives to each agent. | Agentic Governance |
| MCP | Model Context Protocol. A standard interface that agents use to call tools. | Agentic Governance |
| MDSE | Manifest Driven Secure Execution. The model of manifests, pull requests and engine. | Core Concepts |
| MetaGroup | A virtual graph node that makes a set of resources finish before their dependents start. | Architectural Overview |
| Plan | The list of changes that a run calculates. The platform posts it on the pull request. | Get Started |
| PreRelease | The automatic deployment of one new build through preReleaseStages. | CI/CD |
| Promise | A typed placeholder for a value that another resource has not created yet. | Architectural Overview |
| Pull request | A proposed Git change that people can read before they merge it. Each change to a manifest arrives as one. | Get Started |
| Release | A bundled deployment of an ApplicationSet through releaseStages. | CI/CD |
| Release Candidate (RC) | An immutable image tag, <appSet>-<bundle>-rc.<N>, that a release promotes. | CI/CD |
| Run | One execution of the engine against a commit. | Architectural Overview |
| Shared VPC | The hub network in the Core Project. Service projects are the spokes. | Security Framework |
| State ledger | The append-only record of state in Spanner. | Core Concepts |
| Three-legged authorization | An agent acts for one person, with that person's consent. The agent code asks for it. | Agentic Governance |
| Token Broker | A service that gives CI short-lived GitHub tokens. | CI/CD |
| Workload Identity Federation | The way CI signs in to Google Cloud with OIDC and no stored key. | CI/CD |