> ## Documentation Index > Fetch the complete documentation index at: https://docs.orq.ai/llms.txt > Use this file to discover all available pages before exploring further. # Understanding AI operations in Control Tower > Non-technical guide to monitoring AI agents, tracking costs, and maintaining control over your AI operations with the Agent Control Tower. You're not here to write code. You're here to understand **what your AI agents are doing, what they cost, and whether they're working**. This guide is your non-technical walkthrough of the [Agent Control Tower](/docs/control-tower/overview). *** ## Why this exists As AI adoption scales, agents end up scattered across teams, repositories, and frameworks. Different departments build on different tools. Nobody has one place to see what's running, what it costs, or whether it's approved. The Agent Control Tower solves this. It's a **single source of truth** for your entire AI landscape, whether agents were built on Orq, [OpenAI](/docs/proxy/frameworks/openai#observability), [Vercel](/docs/proxy/frameworks/vercel-ai#observability), [LangGraph](/docs/proxy/frameworks/langgraph#observability), or any other framework. **One dashboard instead of five systems.** *** ## The Live Monitoring Overview Your executive dashboard. Six metric cards give you an instant health check. View the full [Control Tower Overview](/docs/control-tower/overview) for real-time monitoring. Control Tower Overview showing six metric cards with sparkline trend comparisons | Metric | What it tells you | | ---------------------- | ----------------------------------------------------------------------------------------- | | **Active Agents** | How many agents are running vs. registered. Low ratio = adoption or retirement questions. | | **Total Tasks** | Tasks processed in the selected period. Your throughput indicator. | | **Total Errors** | Errors across all agents. Rising errors + flat tasks = something is degrading. | | **Total Tokens Usage** | Raw token volume. Drives cost and shows overall compute load. | | **Total Cost** | What your AI operations cost. The number your CFO will ask about. | | **Avg Cost/Task** | Cost per task. Helps spot efficiency differences between agents. | Each card compares current period vs. last period with a sparkline, so trends are visible at a glance. ### Agents Performance table Below the metrics, every agent in your organization is listed with its name, environment tags, department, cost, invocations, and errors. This is your inventory: one place to answer **"what agents do we have and what are they doing?"** Agents Performance table showing environment tags, department, cost, invocations, and errors *** ## The Assets view Drill deeper across four tabs: **Agents**, **Tools**, **Deployments**, and **Models**. See the [Assets documentation](/docs/control-tower/assets) for detailed information. Assets view with navigation tabs and summary cards The top cards show total cost, total runs, active agents, and a **Top Used Agents** leaderboard. Below, each row shows cost, invocations, errors, endorsement status, and last active timestamp. You can search, filter, sort, and configure columns. Agent row detail showing endorsement badge, department, cost, invocations, and last active *** ## Key concepts **Endorsement** A green "Production" badge means a human has signed off on that agent. Think of it as a **quality gate**. No badge? It may be experimental or unreviewed. Your question should always be: **which production agents are endorsed, and which are running without sign-off?** Agent details showing endorsement badge and metadata **Department tagging** Agents are assigned to departments (e.g., "Engineering"), giving you **cost allocation and accountability**. If one department drives 80% of AI spend, this is where you see it. **Environment tags** Tags like "Production" show where an agent is deployed. Production agents carry different risk than staging or development. The Control Tower makes this visible. *** ## Your weekly checklist Five things to check that keep you informed and ahead of problems. | Check | What to look for | | ------------------------ | ---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | | **Cost trend** | Is the Total Cost period-over-period comparison stable? **Sudden spikes mean something changed:** new agent, looping behavior, or model upgrade. Sort Assets by cost to find the source. | | **Error rate** | Total Errors relative to Total Tasks. If the ratio climbs, drill into the Performance table to find which agent is failing and take it to engineering with a specific ask. | | **Active agents** | Compare active vs. registered. Unrecognized active agents = governance gap. Expected agents that are idle = adoption or reliability issue. | | **Endorsement coverage** | Scan the Performance table for production agents without endorsement. **Every production agent should be signed off.** | | **Top consumers** | Check the Top Used Agents widget. The top agents drive most of your cost and volume. Make sure you know what they do, who owns them, and whether usage is expected. | Top Used Agents widget displaying agent rankings by usage *** ## What's coming next What you see today is the visibility foundation. We're building toward expanded cost and spend analytics, risk and compliance dashboards, and custom views for different stakeholders in your organization. *** ## Getting started If your engineering team has already connected agents, you're ready. Log in, explore the [Overview](/docs/control-tower/overview), and run through the checklist above. If agents haven't been connected yet, share the [Getting Started](/docs/control-tower/getting-started) guide with your engineering team. **Setup takes minutes per agent.** Questions? Reach out to your Orq.ai account team for a live walkthrough.