A
| Term | Meaning |
|---|---|
| agent (AI agent) | An AI program that does work rather than just answering questions: it plans, uses tools against your real systems, and reports back. CloudThinker ships five agents — Alex, Oliver, Tony, Kai, and Anna — each owning a domain such as cost, security, or Kubernetes. |
| agentic loop | The four-stage cycle every module runs — Detect → Analyze → Resolve → Validate — explained in full in How CloudThinker works. |
| AgenticOps | Running cloud operations with AI agents that take real actions under rules you set — the Autonomous Cloud Operations category CloudThinker belongs to. The AgenticOps field guide covers the idea in depth. |
| artifact | A dashboard, report, comparison, scorecard, diagram, or saved generated file an agent creates from your connected cloud data. See Artifacts. |
| automation | Stored agent instructions that run whenever a trigger fires — a schedule, an inbound webhook call, or a repository event. See Automations. |
| autonomy | How much an agent may do without asking you first. A CloudThinker workspace has exactly two modes — Manual, where you approve each agent write action, and Auto, where an AI classifier auto-approves writes. See Auto Mode. |
B–C
| Term | Meaning |
|---|---|
| BYOK | Bring Your Own Key: connect your own AWS Bedrock credentials so agents’ LLM usage runs through, and bills to, your AWS account instead of platform credits. See BYOK. |
| Cluster | In Pulse, a group of related surviving monitoring events treated as one candidate Signal — not a Kubernetes cluster. See Signal Clusters. |
| command | A reusable prompt saved to your workspace. Type / in any chat to run one instead of retyping the instruction. See Commands. |
| connection | An authenticated link between CloudThinker and one of your services — a cloud account, database, or monitoring tool. Connect a service once and every agent in the workspace can use it. See Connections. |
| credential | A workspace secret — API key, token, or password — stored encrypted and handed to agents as environment variables when they run commands. See Credentials. |
| Cyber | The module where Oliver security-tests your application, using a login and your source code, and shows you proof of what he found. See Cyber. |
F–I
| Term | Meaning |
|---|---|
| finding | Something an agent detected that needs a human look, tracked with a status until it is settled. Optimize Findings is the work board for savings findings verified during a scan. |
| human-in-the-loop approval | The safety pattern where an agent pauses a specific action until a person decides. In CloudThinker, Tool Permissions set whether each connection tool always runs, needs approval, or is never allowed. |
| Incident | An operational problem that Resolve investigates — created automatically by Pulse or logged manually from a human report. |
K–M
| Term | Meaning |
|---|---|
| keeper | Legacy: keepers were autonomous monitors that enforced cost, security, and performance guardrails. That surface is retired; Optimize Findings is the current cost-finding workflow. |
| knowledge base | A collection of your uploaded documents and crawled web pages that agents search and reference when they answer. See Knowledge Base. |
| LLM | Large language model — the kind of AI that reads and writes language, and the engine behind each CloudThinker agent. With BYOK you can supply your own LLM credentials. |
| Manual / Auto | The two workspace autonomy modes. See the autonomy entry above. |
| MCP | Model Context Protocol — an open standard that lets AI agents talk to external tools and data in one common way. CloudThinker connections are built on MCP, and CloudThinker can also act as an MCP server for other AI clients. |
| mention | The @agent at the start of a prompt that routes your request to a specific agent — @alex, @oliver, @tony, @kai, or @anna. See CloudThinker Language. |
| model tier | The class of AI model an agent run executes on — Light, Pro, or Ultra — which multiplies the run’s credit charge. See Usage. |
| module | One of CloudThinker’s four product areas — Optimize, Review, Cyber, and Resolve. See the four modules. |
O–P
| Term | Meaning |
|---|---|
| On Duty | The workspace toggle for automatic investigation: whether an eligible Pulse-created Incident starts root-cause analysis on its own. See Resolve Settings. |
| Optimize | The module that reviews cost health, current spend, savings actions, cost drivers, and anomalies for your cloud connections. See Optimize. |
| organization | The top-level CloudThinker account that groups your workspaces, team members, and billing. Every user gets one automatically at signup. See Organizations. |
| Pulse | The Detect stage of Resolve. It receives raw monitoring events, removes known noise, groups the survivors into Clusters, and classifies each into a Signal. See Pulse. |
R
| Term | Meaning |
|---|---|
| recommendation | An agent’s prioritized suggestion for a change, with impact and implementation steps. Ask for one with the #recommend tool tag. See Capabilities. |
| Resolve | The module that turns noisy operational events into investigated Incidents, approved actions, and reusable lessons. See Resolve. |
| Review | The module that reviews every pull or merge request with AI, using context from your running infrastructure and your team’s conventions. See Review. |
| run | One execution of an agent’s stored instructions, started by a trigger or by Run now. Each run keeps its full conversation, so you can read exactly what the agent did. See Runs and Calendar. |
| runbook | A written operational procedure. Resolve can connect your runbooks and propose their commands during an Incident, with mutating commands under your control. See Runbooks. |
S–T
| Term | Meaning |
|---|---|
| side panel | The panel beside a chat where you watch an agent team work — follow each subagent’s task, answer requests that need you, and replay the commands agents ran. See Agent Teams and the Side Panel. |
| Signal | In Pulse, a classified operational problem formed from a Cluster of related events. An actionable Signal can route into an Incident. See Signal Clusters. |
| skill | A package of instructions that teaches agents your environment’s procedures and standards. Agents apply an enabled, attached skill when it is relevant to the task. See Skills. |
| subagent | A temporary helper agent that the agent leading your chat spawns for one bounded task; it does that job and finishes. Watch subagents work in the side panel. |
| tool tag | The #word placed right after an agent mention that names the output you want — #visualize, #dashboard, #report, #recommend, #alert, or a knowledge-base name. See Capabilities. |
W
| Term | Meaning |
|---|---|
| workspace | An isolated environment inside your organization with its own agents, connections, and knowledge bases — one per team, project, or environment. See Workspaces. |
| workspace memory | The context agents retain between conversations — environment notes, connection notes, people preferences, and a daily recap — shared across the workspace. See Workspace Memory. |
Related
Quickstart
Register, connect AWS, and run your first cost analysis
CloudThinker Language
The full
@agent #tool instruction syntax referenceFirst prompts by role
Copy a prompt for your role and get a real result in about 10 minutes
AgenticOps field guide
The concepts and governance behind agent-run operations