Dashboard, usage, analytics, and reports
Read the Agent Spend Ledger, the usage meter, Analytics & Insights, and Executive Reports.
Best for: Finance and operators reading spend, the event meter, and downloaded reports
Dashboard, usage, analytics, and reports
These four screens are in the top nav: Dashboard, Usage, Analytics, and Reports. Each one answers a different question. Dashboard is the current spend picture. Usage is Onaro’s event meter for your plan. Analytics splits patterns, findings, and unmanaged AI. Reports downloads the packets finance and audit ask for.
Dashboard
Open Dashboard. The title is Agent Spend Ledger. The line under it is "Meter, attribute, and prove your AI spend".
The window buttons are MTD, 30D, and 90D. The line under them reads "Showing {window} · {start} → {end}. MoM uses same calendar days." The window names are Month to date, Last 30 days, and Last 90 days. Month-over-month compares the same calendar days, not a full prior month against a partial current month.
While it loads, the page says "Loading intelligence...". A failed load says Error Loading Dashboard and offers Try Again. A stale cache says "Showing cached data (may be outdated). Click Refresh to update."
A usage banner appears only when the meter has something to say. The text is the message from the meter. When overage has accrued, a second line says "Overage accrual: $X.XX". The link is View usage report, which opens Usage. If setup is unfinished, a banner says "Setup is not complete. See what Meridian has seen working and what is still missing." and opens /meridian/setup.
The four cards
| Card | How to read it |
|---|---|
| {window} spend | Dollars in the window. The comparison line is "vs same days last month". An increase is shown in red. A decrease is shown in green. |
| Connected Sources | Count of sources that contributed in the window. The caption is "sources in window". |
| Total Tokens | Token volume in the window. Large numbers use K, M, and B. |
| Next Month Forecast | The projected next month, "aligned with MoM trend". If there is no trend to extend, the value is N/A. |
Spending Alerts lists anomalies for the window.
Findings — opportunities identified appears only when there is a findings card to show, at least one recommended guide, and a potential savings figure above zero. The cap line is "Identified opportunity capped at ≤40% of metered spend". The monthly figure is labeled "identified /mo". Annual opportunity (illustrative) is that monthly figure times twelve. Recommended Guides is the count. View Findings → opens /meridian/implementation-center/guides.
Getting started and quick actions
When no vendor has spend in the window, the page shows getting started:
- Create your company, when a company is not set yet.
- Connect a source. The note is "The 14 days start when the first connection succeeds."
- Implementation guides.
Quick Actions are always available: Agents & rules (the ledger), Add billing data (uploads), and Connect source.
Spend by Vendor
The table is Spend by Vendor · {window}. Columns are Vendor, Spend, Models Used, and Total Tokens.
An empty table says "No sources connected yet", then "Connect a provider, gateway, or billing export and the ledger starts metering within hours.", with Connect a source.
Attribution
The attribution block is titled Attribution. The line under it is "{window} · N agents · N active rules". Manage agents & rules → opens the ledger.
Three cards:
| Card | How to read it |
|---|---|
| Attributed | Percent of the window, and the dollars, tied to an agent. |
| Unattributed | "N% of spend still open". |
| Total in window | "Across N spend records". |
The table columns are Agent, Department, Workflow, and Spend, with the window code on the spend header. The open remainder is a row Unattributed — open inbox →, which opens /meridian/ledger?tab=inbox. An empty table says "No agent spend in this window yet."
A failed summary says "Ledger summary failed ({status})". A failed agent list says "Ledger agents failed". Retry loads it again. While it loads, the block says "Loading agent ledger…".
Usage
Open Usage. The title is Usage. The line under it is "Attribution event metering — reports available for audit at any time." ← Dashboard returns to the dashboard.
While it loads, the page says "Loading usage…". A failed current period says "Current usage failed ({status})". A failed period list says "Periods failed ({status})". A failure that cannot be read says "Failed to load usage".
The four cards
| Card | How to read it |
|---|---|
| Events used (MTD) | Attribution events counted so far this month. |
| Included volume | Events included in the plan for the period. |
| % of {basis} | Percent of the alert basis the meter is using. The label is the basis name, not a fixed word. |
| Projected month-end | Where the month is headed if the current pace holds. |
An overage line appears only when overage accrual is above zero: "Estimated overage accrual this period: $X (billed monthly in arrears). Resets {reset_date}."
A hard cap says "Hard cap active — ingest suspended until {reset_date}." New events stop until that reset date. The cards still show what was already counted.
Period statement
The period menu lists "{period} (current)" and each closed period. Download CSV saves meridian-usage-{period}.csv. A download that is not accepted fails quietly and does not change the page.
The statement lines are Events total, Included, Billable, Rate / 100K, and Amount. Rate / 100K is the price per 100,000 events, shown in dollars (cents on the statement divided by 100). Amount adds " (estimate)" when the period is still an estimate.
Daily bars show events by day. An empty period says "No events in this period."
Closed periods are a table with Period, Total, Billable, and Amount. Click a row to select that period and load its statement.
Analytics
Open Analytics. The title is Analytics & Insights. The tabs are Analytics, Meridian Insights, and Shadow AI. The address uses ?tab=insights or ?tab=shadow-ai. The Analytics tab clears the query.
| Tab | Line under the title |
|---|---|
| Analytics | Deep dive into your AI spending patterns |
| Meridian Insights | Meridian Insights — findings and monitoring |
| Shadow AI | Unmanaged AI discovery from event ingest |
While the page loads, it says "Loading analytics...".
Analytics tab
The time buttons are 7 days, 30 days, and 90 days.
Spending Trends plots the first ten dates in the window. Each point is the day’s total, shown to two decimals.
A failed trend says "Error loading spending trends: {message}". While trends load, it says "Loading trends...". With nothing connected, it says "No sources connected yet" and links to /meridian/integrations.
Unattributed spend and idle capacity are separate. They are not added together.
| Figure | Caption | What it means |
|---|---|---|
| Unattributed | Couldn't tell whose it is · last {N} days | Hover text: "We couldn't tell whose it is: spend Meridian saw but could not tie to an agent." |
| Idle | Nobody used it · last {N} days | Hover text: "Nobody used it: private-inference capacity that was paid for but sat idle." Pool periods opens /meridian/settings/cost-pools. |
Top Models by Spend columns are Model, Vendor, Requests, Total Cost, and Avg/Request. When the model breakdown is not available yet, the page says "This feature is coming soon. Model breakdown data will be available shortly.", then Model breakdown unavailable and "Unable to load model breakdown data".
Spending by Department or Spending by Cost Center appears only when the account includes governance. The buttons are Department and Cost center.
Department columns are Department, Requests, Total Cost, Budget, and Members. Total Cost is red when the department is over budget. A missing budget is an em dash. Empty copy is "No department data available".
Cost center columns use the same spend figures. % of Spend is the share, or an em dash when there is no share. Empty copy is "No cost center data available".
Meridian Insights
Three blocks sit above the hero:
- Findings. While it analyzes, it says "Analyzing your spending patterns...". With nothing material, it says "No material opportunities identified" and "We couldn't find stronger model alternatives for your current usage." When there is a finding, the header is "Identified opportunity (capped)" and a monthly figure.
- Spending Alerts.
- Spending Anomalies.
The hero is Meridian Insights. The paragraph is "Meridian findings using Model Intelligence (345+ models) to surface opportunities identified, set alerts, and catch anomalies before they become expensive." The three captions are Findings ("Cheaper model alternatives with attribution"), Spending Alerts ("Stay within budget limits"), and Anomaly Detection ("Catch unusual patterns early").
Shadow AI
Before event ingest is set up, the tab shows Shadow AI discovery. The paragraph is "Shadow AI surfaces unmanaged AI usage from endpoint and gateway events — discover processes talking to AI destinations, attribute them, and promote them to owned agents in the ledger. It runs on event ingest; there is nothing to show until a key with events:write is active."
The three lines under that are "Census unmanaged AI clients on endpoints", "Review new discoveries and promote to agents", and "Close the loop with chargeback and governance". Open Integrations → Discovery opens /meridian/integrations?section=discovery. Set up event ingest opens /meridian/settings/integrations/event-ingest.
Once ingest is on, the inner tabs are Overview and Sources. Event Ingest settings is on the right. The refresh control is labeled Refresh.
Overview panels:
| Panel | How to read it |
|---|---|
| Blind spot | Attribution coverage as a percent, and Discovered unmanaged sources. The line under the numbers is "Your blind spot, and what's in it." Before any of that exists: "Waiting for spend history and the first discovery event." |
| New discoveries (30 days) | A bar per day. An empty panel says "No data yet." |
| Top destinations | Destination, category (or "uncategorized"), and source count. |
| Status breakdown | New, Acknowledged, Brought under management, and Dismissed. |
| Most recently seen | Host, user, and destination, with how long ago it was seen. |
Sources starts on New. The filters are All, New, Acknowledged, Promoted, and Dismissed.
An empty filter says "No discoveries in this filter yet" and "Event ingest is enabled. First discoveries appear after agents send discovery.ai_usage events." Check ingest health opens event ingest settings.
Each source is "{process} → {destination}". Unknown process reads "Unknown process". A missing destination reads "unknown". The next line is host, user, and category. Then "Status {status} · Seen {n}× · Last {when}".
Acknowledge is on new sources. Promote and Dismiss are on new and acknowledged sources. A promoted source links View promoted agent to the ledger. A failed action says "Action failed." unless the server sent its own message.
Reports
Open Reports. The title is Executive Reports. The line under it is "Ledger-based reports — agent, department, workflow, and attribution".
Findings says "Action guides for cost opportunities, model alternatives, and implementation steps." Open Findings opens /meridian/implementation-center/guides.
Export
Export says "Every data export in one place. Open the source screen to download." The rows are links, not downloads on this page:
| Row | Link |
|---|---|
| Allocation detail (CSV) | Chargeback → Lock / download |
| JE import file (per GL system, zip) | Chargeback → Export as JE |
| QBO JE posting history (CSV) | QuickBooks → JE posting history |
| Usage period statement (CSV) | Usage → Download CSV |
Chargeback itself is covered in the chargeback guide.
Available Reports
Each card ends with "Last generated: {date and time} by {name}", or "Never generated." A status other than complete is appended in parentheses.
| Card | What it contains | Who can generate it |
|---|---|---|
| Agent Spend Report | "Assessment PDF for the CFO read-out: unattributed spend, Attribution Readiness, agents, departments, drift, and capped savings." Badge: All tiers. | Every tier. The generate block says "Available on all tiers · assessment deliverable". |
| Board Pack (Monthly) | "Board-ready monthly PDF: spend by department and agent, attribution status, trends, and savings — vendor detail in appendix." | The card says "Automated monthly: all tiers" and "On-demand: Growth and above". The generate block is Generate Board Pack (Monthly) — On-Demand and says "Growth and above · Automated monthly reports available to all tiers". |
| Quarterly Business Review (QBR) | "PowerPoint: financials, savings identified & captured, attribution progress, agent estate, risk, and recommendations." | The generate block appears when the account includes the QBR deck. |
| Risk & Anomaly Report | "Ledger risks (unattributed spend, ownerless agents, unmatched sources) plus vendor concentration and anomalies." | The generate block appears when the account includes the risk report. |
| Audit Export | "Audit-ready provenance: attribution method, rule, actor, and admin / report generation log." | The generate block appears when the account includes audit evidence. |
| Chargeback / GL Export | "Month-locked CSV allocating AI spend to cost centers, ready for your GL. Unallocated dollars post to suspense — never silently dropped." Badge: Growth+. | Open Chargeback → opens Chargeback. This page does not generate the file. |
A card the account cannot generate on demand shows an upgrade badge. The description on the card stays visible.
Generating a file
Agent Spend, Board Pack, Risk & Anomaly, and Audit Export share a Window menu: Last 14 days, Last 30 days, Last 60 days, Last 90 days, This month, Last month, and Custom. Start Date and End Date sit under it. Changing either date switches the window to Custom.
Agent Spend starts on Last 14 days. Board Pack starts on Last month. Risk and Audit start on Last 90 days.
| Generator | Button | File | While it runs | If it fails |
|---|---|---|---|---|
| Generate Agent Spend Report | Generate & Download PDF | agent_spend_report_{start}_to_{end}.pdf | Generating... | "Failed to generate Agent Spend Report", or "Failed to generate Agent Spend Report. Please try again." |
| Generate Board Pack (Monthly) — On-Demand | Generate & Download PDF | executive_report_{start}_to_{end}.pdf | Generating... | "Failed to generate report", or "Failed to generate report. Please try again." |
| Generate Quarterly Business Review (QBR) | Generate & Download QBR Deck | QBR_Q{n}_{year}.pptx | Generating... | "Failed to generate QBR deck", or "Failed to generate QBR deck. Please try again." |
| Generate Risk & Anomaly Report | Generate & Download Risk Report | Risk_Anomaly_Report_{start}_{end}.pdf | Generating... | "Failed to generate Risk & Anomaly report", or "Failed to generate Risk & Anomaly report. Please try again." |
| Generate Audit Export | Generate & Download Audit Report | Audit_Evidence_Report_{start}_{end}.pdf | Generating... | "Failed to generate Audit Export", or "Failed to generate Audit Export. Please try again." |
The download button stays disabled until both dates are set. The button label while the file is building is Generating....
QBR uses Quarter and Year instead of the window menu. Quarters are Q1 (Jan - Mar), Q2 (Apr - Jun), Q3 (Jul - Sep), and Q4 (Oct - Dec).
Audit Export, before compliance configuration is saved, shows "Setup Required: Complete compliance configuration to generate full audit reports." and Complete Setup (2 minutes). Choosing generate before that setup opens the same setup. If both dates are missing, the page says "Please select both start and end dates".
The tip at the bottom of the page: "Agent Spend Report is the assessment deliverable — available on every tier. Board Pack automated emails still go out on the 1st when subscription is enabled. Each card shows Last generated, or Never generated."