CLAWPOD ENGINEERING
Deciding What to Look At First — the ClawPod Cloud Administrator Dashboard
Align periods and scope, separate activity, resource limits and costs, and use a complete illustrative check report.
This guide is for operations owners who read the ClawPod Cloud administrator screens. You will learn to align reporting periods and scope, distinguish activity from resource limits and costs, and choose what to inspect next. You need access to the relevant organization and menus. This article concerns the Cloud Portal administrator dashboard, not an individual Agent’s execution-environment screen.
Choose the period and question first
The official Guide starts at Administrator → Dashboard in the sidebar → period selection. It lists quick insights, top agents, recent activity, and workspace usage, then directs you to an Agent link or Manage → Agents for further status checks. Using the dashboard and checking costs
The questions differ even when they share a screen. The following is an editorial reading strategy, not a product default order or a definition of its metrics.
- To review activity: use quick insights and recent activity to identify what needs further inspection.
- To inspect an individual Agent: consult top agents, but do not treat ranking alone as proof of business performance or throughput. This article has not verified its aggregation criteria.
- To add people or Agents: check current counts and limits in workspace usage.
- To explain costs: open LLM usage costs under Analytics and compare the same periods and organization scope.
Record the reporting period, organization scope, observation time, and screen name beside each number. If two people are looking at different periods, align them before comparing. Matching scope does not guarantee identical values or causes.
Read resource limits separately from model costs
Workspace usage displays the tenant’s plan and member, organization, and Agent counts and limits. Here, tenant means the scope over which those counts and limits are tracked. This is not the capacity of the file workspace used by an Agent. If planned additions exceed the current limit, record the required count and current limit, then discuss cleanup or a plan change with the tenant administrator. Reading this screen does not authorize expansion or billing changes. Checking usage and Agent status
Read costs at Administrator → Analytics → LLM usage costs. An LLM is a language model that generates responses. This screen shows total cost, costs by model, organization and Agent, and daily spend, among other items. The Guide lists active Agents, processed messages, total cost and tokens, LLM calls, and conversation and organization metrics for the analytics summary. Tokens are units used by models to process text. Menus and data scope depend on roles and permissions. Analytics summary and LLM usage costs
LLM usage costs and subscription payment history in plan management are separate. Keep their sources separate in reports, too. Costs and call counts increasing together is a lead for further investigation, not proof of a cause.
Check unusual numbers against details
The Guide recommends checking activity and error rate in the analytics summary, comparing changes in usage, models and Agents with rising costs, then inspecting Agent details, logs and related webhook events if the cause remains unclear. Do not determine a cause from metrics alone or change settings automatically.
Under Manage → Agents, find the target using search and status and organization filters, then open Details on its card. First check whether it is running, stopped or in error. Detail sections can vary with environment and permissions. When something is wrong, inspect AI resource settings, settings and logs together rather than restarting immediately. Managing Agent status and settings
A webhook is a request that delivers an event from an external service. If a request arrived through a webhook, inspect its processing record at Integrations → Webhooks → Events. If several Agents show the same problem, send the start time and scope of impact to the operations owner.
Stopping or restarting requires Agent modification permission and can affect ongoing work or conversations. For an operational Agent, announce the action in the Room and confirm that current work has finished before acting. A Room is a conversation space for users and Agents. The example in this article allows inspection and reporting only, not changes.
Complete one illustrative check
Illustrative example: the request, numbers and result below are not an actual product screen or operating record, and do not imply automatic collection. They show a report written by a person reading screens they are authorized to access.
Request: “Compare LLM usage costs for the Customer Support organization for September 14–20 and September 21–27 using the same scope. Record the reporting periods, organization, observation time and screen name. Separate confirmed changes from unknown causes. Do not change settings or plans, or restart anything.”
Filled result: “At 09:10 on September 28, I checked the Customer Support organization in LLM usage costs. September 14–20 cost $40 and September 21–27 cost $60, an increase of $20. The periods have the same length and organization scope. The causes of the increase by model and Agent remain unchecked. Next, inspect the cost breakdown by model and Agent, followed by the target Agent’s details and logs. No settings were changed and nothing was restarted.”
Verify success: Compare both periods, organization, currency and amounts with the source screen. Check that the report includes the observation time and screen name, and has not turned an unknown cause into a conclusion. Completion here means a completed comparison report, not a diagnosed cause or completed recovery.
If the check fails: If a menu is missing or scopes cannot be aligned, hold the comparison rather than estimating values. Ask the tenant administrator to check the required usage or status. If several Agents are affected, include the start time and impact. Do not guess missing menus or addresses. Platform operations menus System and Monitoring are outside this customer check procedure.
Bring one reading rule to the next meeting
Try this first team rule: “Align the period and scope, then start with activity, resource limits or costs according to the question, and check unusual signals against detailed evidence.” Name an owner and a check time so the next person can repeat the check under the same conditions.
This article explains procedures in the official Guide. It is not a record of observing a current authenticated product screen; actual menus and scope must be checked against your deployment and permissions. Metric aggregation, refresh intervals and query performance have not been verified. If you separately perform an approved change, do not stop at a successful save: follow the screen’s instructions and check Agent status, a small test request in a Room, logs and metrics.
