The glossary
The vocabulary of agentic platform engineering
A category moving this fast needs definitions that hold still. 23 terms, each with a working definition and a link to the deep page where it lives.
- #Agent
- Software that perceives state, reasons over goals, and takes actions toward them with some independence over how. In this discipline: an agent acts on PLATFORM work, through governed interfaces.→ deep page
- #Agentic AI
- AI systems that plan and act autonomously toward goals — deciding steps, not just answering prompts. The capability family that agentic platforms put to work safely.
- #Agentic engineering
- Building agentic APPLICATIONS — products whose users are served by agents. Distinct from agentic platform engineering, which builds the internal platform agents operate through.→ deep page
- #Agentic platform
- An internal platform whose golden paths are machine-executable contracts that governed agents can operate — the agent-native successor to the IDP.→ deep page
- #Agentic platform engineering
- The discipline of designing software platforms in which autonomous agents can perceive operational state, reason over goals and constraints, and take governed action through platform capabilities — while remaining bounded by policy, identity, observability, and human authority.→ deep page
- #Agent contract
- A golden path expressed as a machine-readable contract: explicit inputs, policy expectations, and validation semantics an agent can execute against programmatically. Intent in, verified state out.→ deep page
- #Agent identity
- A first-class, attributable identity for each agent: scoped, time-bounded, delegable with attenuation, and auditable. The plane that bounds autonomy.→ deep page
- #Agent memory
- Scoped, persistent record of past resolutions, decisions, and outcomes — episodic, semantic, procedural — retrieved as reasoning context so the platform learns.→ deep page
- #Agent observability
- Telemetry for the agent loop: decision traces, policy verdicts, action evidence, and behavioral evals. No invisible autonomous changes.→ deep page
- #AgentOps
- The discipline of operating agents themselves — versioning, evaluating, monitoring behavior and cost. Fleet operations; the platform is the roads and traffic law.→ deep page
- #AIOps
- AI applied to operations telemetry: anomaly detection, noise reduction, correlation. Observes and informs; rarely acts. One perception input to an agentic platform.→ deep page
- #Autonomous remediation
- An agent detecting a problem, planning the fix, and executing it through governed interfaces — the top rung of the autonomy ladder for an action class.→ deep page
- #Context engineering
- The practice of assembling what an agent sees per request: perception data, retrieved memory, constraints. Grounding failures are context failures.
- #Golden path
- The paved-road template a platform offers — codified organizational defaults. In the agentic era, golden paths become agent contracts.→ deep page
- #Harness
- The runtime environment an agent works inside — CLI agent, coding assistant, orchestration framework. The builder lane; swappable by design.→ deep page
- #Human-in-the-loop
- A control pattern where consequential actions pause for explicit human approval before executing. The default posture for irreversible actions.→ deep page
- #Human-on-the-loop
- A control pattern where agents act autonomously and humans supervise via traces, alerts, and escalation — the L4/L5 supervision mode.→ deep page
- #Internal developer platform (IDP)
- The self-service platform layer — portals, templates, golden paths — built for human developers. The substrate an agentic platform evolves from, not an obstacle.→ deep page
- #MCP (Model Context Protocol)
- An open protocol (2024-11) for connecting models to tools and data sources through servers. Standard plumbing for the tool layer agents consume.→ deep page
- #Policy-as-code
- Rules expressed as executable, versioned, testable code (OPA/Rego, Kyverno) that validate every proposed action before mutation. Fails closed; prompts are suggestions, policies are gates.→ deep page
- #Progressive autonomy
- The read → recommend → propose → execute → remediate ladder, climbed per action class on evidence and revocable at any time.→ deep page
- #Sense-Think-Act loop
- The core architectural pattern of agentic platforms: perception ingests state, reasoning plans against goals and policy, action executes through governed interfaces.→ deep page
- #Tool
- A discrete capability an agent can invoke — API, CLI, MCP server. Tools are the action surface; which tools an identity may call is the governance question.→ deep page
These definitions are published as machine-readable DefinedTerm entries in a DefinedTermSet — the same structure AI engines assemble answers from. Missing a term? The canonical definition page anchors the set.
Ready for the leap?
Partner with Adventure On The Wave to build governed, agentic platform capability — architecture, guardrails, and the human authority model to match.
A strategic initiative by Adventure On The Wave