What is the Dashboard?

The Dashboard is the home page of the Runtype web app at use.runtype.com. It summarizes your workspace in one place: your products, how your agents and flows perform in production, what they cost, and how your most recently scored eval suite did.

To connect the Runtype MCP server to your editor or AI assistant, click Connect your assistant in the page header. If you chose Improve what I already built during onboarding, the button reads Connect your editor. For setup details, see Connect Runtype to coding agents.

The sections that follow describe the page from top to bottom.

Attention

When Runtype detects a warning or critical change in your production metrics, an Attention strip appears at the top of the page. The strip stays hidden when there is nothing to report. It shows up to five notices, most severe first. The first notice is always visible. Expand N more notices to see the rest.

A notice appears for these conditions:

• The success rate fell by 5 percentage points or more compared with the previous period.
• p95 latency rose by 50 percent or more compared with the previous period and is at least 1 second.
• Model cost per successful run rose by 25 percent or more while run volume stayed within 25 percent of the previous period.
• The eval pass rate dropped by 2 percentage points or more compared with the previous scored run.
• At least 10 percent of agent runs reached the turn limit, a higher share than in the previous period.
• At least 20 percent of runs fell back to a different model.
• No runs happened in the selected period, although runs happened in the period before.

The success rate, latency, and cost notices need at least 20 runs in both the selected period and the previous period. The turn limit notice needs at least 20 agent runs in the selected period, and the fallback notice needs at least 20 runs.

Select a notice to open the related Logs, Evals, Agents, or Models page.

Your products

Your products shows your three most recently created products. Each card shows the product’s description and its number of capabilities and surfaces. The surface count shows how many surfaces are configured. It does not show deployment status.

Each card suggests a next step. The label is Add capabilities when the product has no capabilities, Add a surface when it has no surfaces, and Open product otherwise. Click a card to open the product.

To create a product, click Create product. To see every product, click View all. A new workspace with no products shows a Bring your first product to life panel instead of cards, and View all does not appear.

Time range and refresh

Select Last 7 days, Last 30 days, or Last 90 days from the time range menu. The selection applies to the Attention notices, production health, execution costs, and activity.

To reload the metrics, click the refresh button next to the time range menu. The time of the most recent update appears beside it. If a refresh fails, a Couldn’t refresh message with a Retry button appears above the metrics, the timestamp changes to Showing saved data, and the metrics from the previous successful load stay visible.

Production health

Production health opens with a one-line summary of the period and a count of agent and flow runs. Four metric cards follow:

• Runs: top-level agent and flow runs, including runs that have not finished. Steps and sub-agent calls inside a run are not counted separately.
• Success rate: successful runs as a share of finished runs.
• p95 latency: the duration that 95 percent of finished runs complete within.
• Model cost per success: model cost for the period divided by the number of successful runs. The card shows no value when some model costs are missing.

Each card compares the selected period with the previous period of the same length. A green trend line marks an improvement, and an amber trend line marks a regression. A card with no earlier data shows No prior period to compare.

Click a card to open Logs filtered to the same time period. The Success rate card opens the failed runs.

Execution costs

After the metric cards, the cost breakdown shows estimated Model cost, Tool cost, and Total cost for the period. Each run’s nested work counts once. The estimates exclude billing margin, so they do not match your invoice. For billed usage, see Usage data.

When some charges were not reported or could not be priced, each cost shows a known subtotal and the number of runs with incomplete cost data. A Partial cost data note explains that the totals include known costs only. Incomplete cost data does not mean that a run is still running, and a missing Model cost per success value does not mean that the runs were free.

Activity

Activity charts the number of runs per day for the selected period. Select Total, Agents, or Flows to change which runs the chart counts. Days are in UTC. Click a day on the chart to open the logs for that day. If some older runs have no recorded type, a note says that they are included in Total only.

To see the numbers as a table, expand View daily counts. Each date in the table opens the matching logs. To open the runs for the period in Logs, click Explore runs. The Total, Agents, or Flows selection applies to the logs that open.

Product quality

Product quality shows your most recently scored eval suite, its pass rate, the date of the run that produced the score, and a comparison with the previous scored run of the same suite. The pass rate describes that one suite. The scoring date can be earlier than the selected time range. To open your suites and runs, click View evals.

Before any suite has scored, the card shows Set up quality checks with one of these messages:

• No eval runs yet when no eval has run.
• Last run DATE has not scored yet when a run exists but has no score.

If a newer run exists that has not scored yet, the date line ends with newer run not scored yet, and the card still shows the older score.

Product updates

Product updates lists when each product in Your products was last saved. A timestamp here records a save to the product’s details. It does not record a deployment or a change to an agent or flow.

When nothing has run yet

If your workspace has no runs in the selected period or the period before it, and no scored evals, Production health shows a Nothing has run yet panel in place of the metrics. The panel has three actions: Run an agent, Run a flow, and Set up evals.

Troubleshooting

Each area of the page loads separately, so one failure does not hide the rest of the page.

Next steps

• What are Products?: group capabilities and surfaces into a product.
• Daily executions: understand run counts.
• Working with Logs: inspect execution details.
• What are Evals?: set up the suites that feed Product quality.
• Usage data: review billed usage.
• Onboarding credit rewards: earn starter build credit by completing onboarding tasks.