Section Front

Glossary

The vocabulary of agentic systems — concise definitions of the terms that show up in every Triadive article. Each entry stands alone and is referenced from concept, workflow, and lesson pieces. Read these once, then refer back as needed. The glossary is a working reference, not a sequential read. Start with the foundational primitives — loop, graph, memory, session — and branch out from there.

Every Triadive concept, workflow, and lesson assumes a working vocabulary. The glossary entries below are that vocabulary — concise definitions of the terms that recur in the rest of the manual, written so they can be read once and referred back to.

The terms are ordered by category rather than alphabetically. Start with the foundational primitives (loop, graph, memory, session) and branch outward as your work demands it. Each entry is cross-linked from the concept, workflow, and lesson pieces where the term first appears in context, so the manual reads cleanly without requiring you to leave the page you're on.

The glossary is a working reference, not a sequential read. Treat it like a glossary is meant to be treated — a tool you open when a term arrives that you want to nail down, and a tool you close when the term is clear.

Glossary

The vocabulary of agentic systems — concise definitions of the terms that show up in every Triadive article. Each entry stands alone and is referenced from concept, workflow, and lesson pieces. Read these once, then refer back as needed. The glossary is a working reference, not a sequential read. Start with the foundational primitives — loop, graph, memory, session — and branch out from there.

JobA specific assignment given to a work desk: a unit of work with a name, inputs, a deadline, and an expected output. The job is to the desk what a task is to a s Specialty AgentA work desk tuned for a narrow function: a specific role, a specific toolset, and a specific scope of work. A specialty agent is a work desk with a defined spec Work DeskA durable operating context where a named line of work lives: a goal, a role, a memory surface, selected tools, a job queue, and rules for escalation. In a heal WorkboardA card-based state-tracking system for agent operations. A workboard card holds status, assignee, priority, and an audit trail — the system of record for what i CompactionWhen the context window fills, the agent has to summarize older turns into a smaller representation while preserving the facts that matter for the goal. Compact EmbeddingAn embedding is a list of numbers that represents a piece of text so that semantically similar texts are close together in vector space. Embeddings are how sema HITL (Human-in-the-Loop)A pattern where an agent pauses at a defined checkpoint and waits for a human to approve, edit, or veto before it continues. HITL is how an agent system keeps i MCP (Model Context Protocol)MCP is an emerging open standard for connecting LLM-based agents to external tools, data, and prompts. Servers expose capabilities; clients connect to them. The ObservabilityThe practice of capturing enough structured signal from a running agent system to debug it after the fact. Three signal types: traces, metrics, and logs. The di Agent LoopThe repeated cycle an agent runs to make progress on a goal: intake → context assembly → model inference → tool execution → persist → repeat until a stop condit Agentic RAGRetrieval-Augmented Generation where the agent decides what to retrieve, when to retrieve, and how to use the retrieved information. The retrieval is part of th AlgorithmThe decision procedure the agent uses to choose the next action. The algorithm is what turns the agent loop into a policy. Context vs. MemoryContext is what the model sees in the current inference call; memory is what persists across inferences. Context is bounded; memory is unbounded. Cron vs. HeartbeatCron fires at scheduled times; heartbeats poll regularly and decide what to do. Cron is precision; heartbeats are awareness. CronA scheduled trigger that fires an agent loop at a specified time or interval. Cron is the operator's way of making the agent do work on a schedule. Graph-Based MemoryA memory layer where the agent's persistent state is stored as a graph of nodes and edges. The graph is the memory; the retrieval is graph traversal. MemoryThe persistent store that lets an agent pick up where it left off across runs, sessions, and reboots. Memory is what turns a single prompt into a long-running r SandboxA bounded execution environment where the agent's actions are scoped to a defined set of resources. The sandbox is the agent's blast radius. SessionA bounded unit of agent work that has a beginning, an end, and persisted state across multiple loops. Sessions are the units that cron, monitoring, and audit op SkillA versioned, reusable procedure that an agent can invoke on demand. Skills are the manual's equivalent of a function library: prompts, scripts, and conventions State GraphA network of nodes (steps) and edges (transitions) that defines the possible paths and handoffs a long-running system can take. Loops decide; graphs route. ToolA function the agent can call: a piece of code that does one thing well, has a clear input/output contract, and is the agent's interface to the outside world.