Glossary

Same terms as in Introduction → Concepts, plus operator-specific terminology and cross-links to where each concept is configured or used.

Identity & access

TermDefinitionWhere it appears
IdentityA unique entity (user, agent, runner, app) with a platform-wide UUID.Operate → Identity
UserHuman identity, provisioned on first OIDC login.Operate → Identity
AgentAI identity, configured as an Agent resource and run as a workload on demand.Administer → Agents
RunnerIdentity that hosts workloads. Cluster- or org-scoped.Administer → Runners, Operate → Runners
AppIndependently deployed service that participates in conversations as itself.Administer → Apps, Build & extend → Apps
OrganizationMulti-tenant boundary grouping users, agents, models, secrets, runners, apps.Administer → Organizations
Cluster adminPlatform-wide administrative role.Self-host install → First admin, Administer → Cluster administration
Organization ownerPer-organization administrative role.Administer → Organizations
Organization memberNon-owner participant. No Console access.Administer → Members
Agent rolePer-agent grant: owner, maintainer, participant.Administer → Agent roles
API tokenLong-lived credential for programmatic access. Format agyn_<44 chars>.Use → API tokens
OIDC subjectThe IdP-issued sub claim that uniquely identifies a user.Operate → Identity

Conversation

TermDefinitionWhere it appears
ConversationPersistent exchange between participants. Lifecycle: open → resolved.Use → Chat
ThreadThe storage record behind a conversation.Use → Chat, Operate → Architecture
ChatThe platform's user-facing communication interface.Use → Chat
ParticipantUser, agent, or app in a conversation.Use → Chat
ReminderAgent-scheduled follow-up.Use → Reminders, Administer → Reminders app
Conversation statusUser-controlled state: Open or Resolved.Use → Chat
Activity statusSystem-derived state: Running, Pending, Finished.Use → Chat
Degraded threadUnrecoverable state — read-only conversation.Use → Chat

Agent

TermDefinitionWhere it appears
Agent availabilityinternal (any org member) or private (role-restricted).Administer → Agents
Runtime imageContainer image the agent CLI runs in.Administer → Agents
Init imageContainer image that bootstraps agynd and the agent CLI.Administer → Agents, Build & extend → Agent CLIs
Idle timeoutDuration of CLI silence after which the workload stops.Administer → Agents
MCP serverModel Context Protocol server providing tools. Sidecar in the agent pod.Administer → MCP servers, Build & extend → MCP servers
ToolA capability exposed by an MCP server.Build & extend → MCP servers
SkillPrompt fragment placed on the agent's filesystem.Administer → Skills
HookEvent-driven sidecar.Administer → Hooks
VolumePersistent disk attached to an agent or sub-resource.Administer → Volumes
Init scriptShell script run before the agent CLI starts.Administer → Init scripts
Environment variablePlain or secret-backed value injected into a container.Administer → Environment variables

Run & observability

TermDefinitionWhere it appears
RunA single execution cycle of an agent processing unacknowledged messages.Use → Run Timeline
Run eventA discrete step within a run (message, LLM call, tool execution, summarization).Use → Run Timeline
WorkloadA running agent pod.Administer → Monitoring, Operate → Runners
ContainerIndividual container within a workload.Administer → Monitoring
ContextItems assembled into a prompt for an LLM call.Use → Run Timeline
SummarizationRun event compressing context to stay under token limits.Use → Run Timeline
TraceRecorded sequence of events for a run.Use → Run Timeline
SpanA single unit in the trace; OTLP-style.Operate → Architecture

Models & secrets

TermDefinitionWhere it appears
LLM providerConfigured upstream model provider (OpenAI, Anthropic, etc.).Administer → LLM providers
ModelPlatform-side identifier (gpt-4o) mapped to a provider and remote model name.Administer → Models
SecretCredential value, stored locally or by reference to an external provider.Administer → Secrets
Secret providerExternal secret store the platform reads at workload start.Administer → Secret providers
Image pull secretRegistry credentials for pulling private images.Administer → Image pull secrets

Infrastructure

TermDefinitionWhere it appears
GatewayExternal API entry point speaking ConnectRPC / gRPC.Build & extend → Gateway API, Operate → Architecture
OpenZitiZero-trust overlay network for agents, apps, runners, devices.Operate → Networking
IstioService mesh for in-cluster mTLS and authorization.Operate → Networking
OpenFGAReBAC engine backing the Authorization service.Operate → Authorization
DevicePersonal endpoint enrolled into OpenZiti.Use → Devices
Port exposureReachable endpoint for a service inside an agent container. URL http://exposed-<id>.ziti:<port>.Use → Port exposure
NotificationReal-time event delivered to UIs via WebSocket.Operate → Architecture

Components & repositories

TermDefinition
agyndWrapper daemon running inside every agent pod. Bridges any agent CLI with the platform.
agyn CLIPlatform CLI for interactive and scripting access.
agn CLIAgyn's native agent loop implementation.
CodexOpenAI's agent loop, supported via agent-init-codex.
Claude CodeAnthropic's agent loop, supported via agent-init-claude.
files-mcpPlatform-provided MCP server exposing read_file.
Console appBrowser management UI.
Chat appBrowser chat UI.
Tracing appBrowser run-inspection UI.
k8s-runnerDefault Kubernetes runner implementation.
bootstrapInstall path — agynio/bootstrap. k3d cluster + Terraform stacks for every platform service.
platform-chartsCentralized umbrella Helm chart at agynio/platform-charts. In preparation; will replace per-service deployment in bootstrap once stable.
terraform-provider-agynTerraform provider for the Gateway API.
agynio/apiProtobuf schemas for every service.