Glossary
Agent Serve terms, in plain language
Short definitions of the words we use across Agent Serve, each linked to the page that explains it in full.
A glossary of Agent Serve terms, the AI workspace from Valnora & Elric where teams build, deploy, and manage AI agents: Relays, Chat, the workflow builder, the semantic model, rules, knowledge bases, monitors, row-level security, zero data retention, self-hosting and SOC 2 readiness.
Terms
From AI workspace to zero data retention.
- Agent Serve
- The AI workspace from Valnora & Elric where teams build, deploy and manage AI agents. Available now by invitation. Learn more
- AI workspace
- One place where people and AI agents work on the same data, tools and permissions, instead of separate point tools.
- Relay
- An AI coworker in Agent Serve. A Relay has a role, instructions, the resources it may use and per-action permissions, and asks a person before external actions such as sending an email. Learn more
- Chat
- The conversational surface of Agent Serve. Ask questions of your data and documents, see the sources behind answers, and build resources by describing them. Learn more
- Workflow builder
- The visual builder in Agent Serve. Connect triggers, AI agents, code, approvals and integrations into steps you can run, inspect and deploy. Learn more
- Semantic model
- A shared description of business data: the entities, how they relate, and the measures that matter, so dashboards, Chat and agents join and measure data the same way. Learn more
- Rule
- A reusable piece of business logic, such as a flag, a classification or a score, that dashboards, tables and monitors apply the same way. Learn more
- Knowledge base
- Documents and synced sources indexed so agents can find the passage that answers a question and show where it came from. Learn more
- Monitor
- A check on a metric or condition that raises an alert when it crosses a line; a Relay can pick up the follow-up. Learn more
- Row-level security
- Rules that decide which rows of data each person or team can see, applied in tables and dashboards. Learn more
- Zero data retention
- Running models under a provider’s terms that do not store prompts or outputs. In Agent Serve this means your own provider keys and agreements, or self-hosted models. Learn more
- Self-hosting
- Running Agent Serve in your own cloud account or VPC with its Helm chart or Docker images, instead of our cloud. Learn more
- SOC 2 ready
- The controls a SOC 2 audit looks for are in place. It is not a certification or an audit report. Learn more