Use a project's Quality coverage tab to see which data assets your agents use, which of those assets have running data quality (DQ) policies, and how to get policies added where they're missing.
Prerequisites
- Your project has traces that read or write data. Assets show up on the tab only after a trace reads or writes data through an instrumented driver, or after your code declares them with
declare_assets. For how to set that up, see Record the data your agent uses. - You can view the project. The tab appears to users who can view the project. For details, see Project permissions.
Open the Quality coverage tab
- Open your project.
- Go to the Quality coverage tab.
The tab shows the last 7 days by default. To change the range, use the time range. For how it works, see Find your way around.
The tab ranks assets by what instrumentation captured. Data your agent reaches through a driver that isn't instrumented, through an MCP tool, or from code that doesn't declare its assets doesn't appear here.
The page shows a four-tile funnel at the top, an optional filter chip, and then the asset table.
Read the trace funnel
The funnel counts the traces in the selected range in four steps. Each tile counts a share of the tile before it, and some tiles also have a gap line for the traces that dropped out.
| Tile | What it counts | Gap line |
|---|---|---|
| Traces | Every trace in the range. The count matches the traces list. | — |
| Using assets | Traces that called tools or read data. The rest are chat only. | use no assets |
| Assets correlated | Traces whose assets are all identified and matched to an asset in the Acceldata (ADOC) catalog. | not identified · not in ADOC |
| Assets covered by DQ | Correlated traces whose assets have DQ policies that are running. This is coverage as of now, not whether the data was good when the agent read it. | with no running DQ |
A policy counts as running when it's enabled, it's scheduled, and its last run was within the last 24 hours.
Each gap line points to its own fix:
- not identified: the trace called tools, but AIO didn't capture which data it used. Fix this in your instrumentation or with
declare_assets. See Record the data your agent uses. - not in ADOC: the trace used assets that weren't matched to an ADOC asset. Fix this by registering the data source in ADOC.
- with no running DQ: the trace's assets are in ADOC but lack running DQ policies. Fix this by adding policies. See Add quality checks with ADM.
Open matching traces from the funnel
Some tiles open the traces list, filtered to the traces they count. The list keeps the same time range.
- Select Traces to open every trace in the range.
- Select Using assets to open the traces that used assets.
- Select use no assets to open the chat-only traces.
- Select not identified to open the traces that called tools but referenced no data.
For how to read and filter the traces list, see Explore traces.
Filter the asset table from the funnel
Other tiles filter the asset table on this page. Each one applies the following filter.
| Select this tile | Filter applied |
|---|---|
| Assets correlated | Assets in ADOC |
| not in ADOC | Assets not in ADOC |
| Assets covered by DQ | Assets covered by DQ |
| with no running DQ | Assets in ADOC with no running DQ |
A filter chip shows the active filter and how many traces read the matching assets, for example "Assets in ADOC · read by 12 traces". To clear the filter, select the same tile again, or select the remove button on the chip.
Read the asset table
The table lists the data assets the project's traces used, with their quality coverage.
| Column | Shows |
|---|---|
| Asset | The asset name. Hover over it to see its namespace. |
| Data source | The data source name, or the source type when there's no name. |
| Traces | How many distinct traces used the asset. |
| Sessions | How many distinct sessions used the asset. |
| Last used | The date and time an agent last used the asset. |
| Access | read + write if any trace wrote to the asset, otherwise read. |
| Coverage | The asset's coverage state. See the following table. |
| Policies | How many policies the asset has and how many are running, for example 3 · 2 running or 3 · none running. Shows none when the asset has no policies, and — when coverage is Not in catalog or Unknown. Hover over it to see which columns the policies check. |
| Last result | The worst latest result across the asset's policies, ranked failing, error, warning, then passing, and when the last run happened. |
| Columns not checked | Columns your agents used that no policy checks. Column names are compared without regard to case. A policy on the whole table doesn't count as checking any column. |
| Actions | Add quality checks with ADM, on every row whose coverage isn't Checked. |
AIO decides each asset's coverage state by checking these conditions in order and using the first one that applies.
| Coverage | When it applies |
|---|---|
| Not in catalog | No catalog asset matched the reference. |
| Unknown | The match isn't exact, or the asset's policies haven't been fetched yet. |
| Not checked | The asset has no policies. |
| Checked | At least one of the asset's policies is running. |
| Has policies, not running | The asset has policies, but none of them is running. |
Sort and page the table
The table is sorted by Traces, highest first, by default.
- To sort by another column, select its header. Select it again to reverse the order, and a third time to go back to the default sort. You can sort by every column except Last result, Columns not checked, and Actions.
The table shows 25 rows per page. When you change the time range, filter, sort, or page size, the table goes back to the first page.
Add quality checks with ADM
For any asset that isn't Checked, AIO prepares a prompt that describes how your agents use the asset and what its current policies are. You paste the prompt into ADM chat, and ADM prepares data quality, reconciliation, schema drift, data drift, and anomaly policies for you to review.
- In the asset's row, under Actions, select Add quality checks with ADM. A side pane opens, titled with the asset name, for example "Data reliability policies for orders.public.customers".
- Review the prompt in the pane.
- Select Copy prompt. The button reads Copied to confirm.
- Select Open ADM chat ↗. ADM chat opens in a new tab.
- Paste the prompt into ADM chat, and review the policies ADM prepares.
The prompt describes the asset's use in the time range selected on the page. It includes these items:
- The asset's name and data source, and a
source: aiotag for ADM to add to every policy it creates. - How agents use the asset: the number of traces and sessions, a per-day rate, and whether agents only read the asset or also wrote to it.
- The columns agents used and how they used them, such as selected, filtered, joined, grouped, ordered, or written. The prompt lists up to 20 columns.
- Suggested policies, such as data quality, schema drift, freshness, data drift, profile anomaly, and reconciliation.
- Gaps: columns that agents use but no policy checks, and policies that need attention, for example because they're disabled, not scheduled, never ran, failed on their last run, or haven't run in more than 24 hours.
- Up to five example queries, with literal values replaced by
?. - Up to five recent questions people asked the agent, marked as data for ADM to read, not instructions to follow.
Tip: Remove anything from the prompt you don't want to share before you paste it into ADM chat.
Troubleshooting
- "No data access captured in this range": no trace in the range read or wrote data through an instrumented driver or declared its assets. Choose a longer time range, or see Record the data your agent uses.
- "No asset matches this filter.": the active filter matches no assets. Remove the filter chip, or select the tile again.
- "No project called “checkout-agent”": the project in the link doesn't match any project in your tenant. It might have been deleted, or it might belong to a tenant you aren't signed in to. Select Pick a project to choose a project.

Have a suggestion?