> For the complete documentation index, see [llms.txt](https://docs.finout.io/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://docs.finout.io/user-guide/optimize/agents-beta.md). # Agents \[Beta] An Agent continuously scans your cloud infrastructure against filters and metrics you define, and surfaces the resources that match as findings. Instead of running a one-off scan, an agent keeps watching — against your own thresholds — so waste doesn't have to be rediscovered every time someone remembers to look. ### Enable agents for your account {% hint style="info" %} To use the agentic product, enable the AI features in your account. See [AI Opt-In](https://docs.finout.io/cross-platform-features/list-of-cross-platform-features/ai-in-finout/ai-opt-in). {% endhint %} **Permissions** Access to agents is controlled by account-wide permissions: **View Agents**, **Create Agents**, and **Edit Agents**. They're granted to admins by default, and you can assign them to other users as needed. See [Permissions List](https://docs.finout.io/settings/role-based-access-control-rbac/managing-roles/permissions-list). ### How an agent creates findings Every agent is built around a funnel: you choose a **provider**, which narrows the **services** you can pick from; you choose a service, which narrows the **use cases** available for it; you choose a use case (for example, rightsizing or idle), which narrows the **metrics** relevant to detecting it. Every field downstream only shows options that are actually valid given what you've already selected upstream. For example: selecting AWS as the provider shows only AWS services (EC2, RDS, FSx, etc.); selecting RDS as the service then shows only use cases valid for RDS, such as rightsizing; selecting Rightsizing then shows only the metrics relevant to rightsizing RDS, such as CPU utilization. When an agent's rules match a resource, that match becomes a finiding — visible in Findings, with the resource, the estimated savings, and its status. ### How often agents run Agents run on a daily basis. If you create or edit an agent today, it won't show findings until its first run completes — check back the next day. ### Supported services, use cases, and metrics #### Use case definitions * **Rightsizing** — the resource is running but sized larger than its actual workload requires. * **Idle** — the resource shows no meaningful usage over the lookback window. * **Extended Support** — the resource is incurring vendor extended-support fees for running an outdated or end-of-life version. #### AWS | Service | Use case | Metrics | | ---------------------------- | ---------------------------------------------- | ------------------------------------------------------------------------------------------------- | | EC2 — GPU to General Purpose | Rightsizing | GPU utilization, CPU utilization, GPU memory used | | EC2 — GPU to CPU Instance | Rightsizing | GPU utilization, CPU utilization, GPU memory used | | EC2 — GPU to Memory Instance | Rightsizing | GPU utilization, CPU utilization, GPU memory used | | EC2 | Rightsizing | CPU utilization | | EC2 | Idle | CPU utilization, network in, network out | | EC2 — GPU | Idle | CPU utilization, GPU utilization, GPU memory utilization, network in, network out, GPU power draw | | RDS | Rightsizing | CPU utilization | | RDS | Idle | Database connections | | RDS | Extended Support | Billing line-item description | | ElastiCache | Idle | CPU utilization | | Gateway Load Balancer | Idle | New flow count | | Network Load Balancer | Idle | New flow count | | Classic Load Balancer | Idle | Request count | | Application Load Balancer | Idle | Request count | | S3 | Idle | Total requests | | EBS | Idle | Volume status (available/unattached) | | EBS — gp2 | Rightsizing | Usage type (gp2 volumes) | | EKS | Extended Support | API operation | | Elastic IP | Idle | Usage type (idle address), 30-day lookback | | AmazonFSx | Rightsizing — Underutilized Provisioned IOPS | Provisioned IOPS mode, IOPS utilization | | AmazonFSx | Rightsizing — ONTAP Inefficient Volume Storage | File system type, storage efficiency setting | | DocDB | Extended Support | Billing line-item description | #### GCP | Service | Use case | Metrics | | ------------------------ | ---------------- | -------------------------------------------------------- | | VM (Compute Engine) | Rightsizing | CPU utilization | | VM (Compute Engine) | Idle | CPU utilization, received bytes, sent bytes | | Cloud SQL | Idle | CPU utilization, memory utilization, read ops, write ops | | Cloud SQL for MySQL | Extended Support | SKU description | | Cloud SQL for PostgreSQL | Extended Support | SKU description | | Snapshot | Idle | Resource labels (waste type, product type) | | Persistent Disk | Idle | Resource labels (waste type, product type) | | | | | ### FAQs **Can I monitor a service or provider that isn't listed here?** Not yet. Coverage expands over time — this page reflects what's supported today. Have a particular service, provider or use case in mind? reach out to your CSM. **Does the lookback window change per use case?** You can configure the lookback to fit it to your own needs. **What's the difference between Idle and Extended Support?** Idle flags a resource that isn't being used. Extended Support flags a resource that is being used, but is incurring extra vendor fees because it's running an outdated version. ### Where to go next [Managing Agents](https://docs.finout.io/user-guide/agentic-product-beta/agents-overview-beta/managing-agents-beta) --- # Agent Instructions This documentation is published with GitBook. GitBook is the documentation platform designed so that both humans and AI agents can read, navigate, and reason over technical content effectively. Learn more at gitbook.com. ## Querying This Documentation If you need additional information that is not directly available in this page, you can query the documentation by asking a question. Perform an HTTP GET request on the following URL with the `ask` and `goal` query parameters: ``` GET https://docs.finout.io/user-guide/optimize/agents-beta.md?ask=&goal= ``` `ask` is the immediate question: it should be specific, self-contained, and written in natural language. `goal` is what the user is ultimately trying to achieve, the reason they need the answer. Sharing it helps GitBook give you a better, more relevant answer. A goal is most helpful when it describes the outcome the user wants rather than restating the question. For example, with `ask=how do I create an API token`, a goal like `automate deployments from our CI pipeline` lets GitBook tailor the answer to that use case. The response will contain a direct answer to the question and relevant excerpts and sources from the documentation. Use this mechanism when the answer is not explicitly present in the current page, you need clarification or additional context, or you want to retrieve related documentation sections.