Read the dashboard
The Credits page shows the balance, the low balance alert setting and the history of top-ups. Each top-up row has the date, the amount, the credits with any bonus, the payment method and the status. A top-up paid by card or PayPal has a Receipt link that opens its documents from Stripe, the payment processor, in a new tab. Since November 7, 2026, each such top-up also has a paid invoice. The link then opens the invoice page, where you download the invoice PDF and the receipt. Stripe also emails both to the address used for the payment. The invoice names the seller, Artem Zhehlov, a sole proprietorship registered in Poland, and charges no VAT under the Polish VAT exemption for small businesses. Crypto top-ups have no Stripe documents. A crypto top-up appears in the history as soon as you send the coins, with the status Processing, while the network confirms the transfer. Its credits arrive when the payment is confirmed, and the status changes to Succeeded. The Metrics page covers a period from 15 minutes to 30 days. Periods in days count whole UTC calendar days, today included. The totals at the top add up across every model type: spend, requests, error rate and the number of models used. Below them, Spend by type stacks spend by chat, image, video and the other types, and Requests by outcome splits requests into successes and errors. Each type then gets its own tab, the costliest first. A tab shows that type’s spend, requests, error rate and latency, with time to first token where it is measured. Chat adds a tokens chart. Usage by unit lists how much of each billed unit the type used and what it cost. Units are text tokens, images, seconds of video or audio, characters of speech and search calls. At the bottom, spend is split by model, API key and model publisher. Select a model or a key to open its requests on the Logs page. With several workspaces, the account’s Metrics also split spend by workspace. With keys from more than one person, they split it by member too. A request with a key counts in the workspace the key is in now and for the person who created the key. A key with no person behind it counts as No owner. That is a workspace key, a key whose creator left, or a key made before keys recorded who created them. The Playground sends requests without a key, so they count in the workspace they were made in and show as Playground in the member list. You can filter the whole page by API key, model, workspace and member. The Logs page lists every request of the last 31 days, newest first, with every kind of model in one table. A row has its time, the model with an icon for its kind, usage, duration, cost and API key. Click a row to open its details, or move to it with Tab and press Enter or Space. A request that did not succeed carries a tag before the model. Error means it failed, and Canceled means your client closed the connection. Cut off means the model stopped early, for example at the output token limit. Usage is in the unit the model bills. The left of the arrow is what you sent, and the right is what the model returned. That is tokens for a text model, characters for speech, images for an image model and seconds for video. Hover over it for each billed unit. Filters above the table narrow the list by period, kind, model, key, workspace, member, outcome or request ID. The workspace filter shows on the account’s Logs, and the member filter shows once keys from more than one person are in the list. The page address keeps them, so you can share a filtered list. A chart above the table shows the requests of the period over time, colored by outcome. Hover over a bar for its counts. A video that is still being generated has no row yet. It shows in an In progress strip above the table, with its stage, length, resolution and time so far. When it finishes, it leaves the strip and its row appears. Details open in a panel on the right, and the page address keeps the open request, so you can share a link to it. Arrows at the bottom of the panel step to the previous or next request on the page. A streamed answer shows the time to the first token and the output speed, and Timing draws both phases as bars. A section for the model’s kind shows what the request asked for, such as the image size or the video resolution. How the cost was calculated opens to list each billed unit with its quantity, unit price and amount, which add up to the cost. When a video costs more than the price NagaAI quoted at submission, a Price cap line takes off the difference. With request logging on in the workspace, Content shows the response, the images or the inputs. For a video, it shows the prompt and plays the finished clip for the 7 days NagaAI keeps the file. A one-line prompt shows in place. A longer conversation opens from Prompt in a large window, where you can search the messages, filter them by role and copy one. Download JSON saves the raw request and response. Export downloads up to 10,000 rows of the list as filtered, as CSV or JSON. Each log row matches thex-request-id header of the response it describes, so you can find a request from your own logs.
Get the balance with the API
Both account endpoints take a provisioning key. To read the balance, callGET /v1/account/balance.
Get a spending summary
GET /v1/account/activity?days=N summarizes the last N days, from 1 to 30, and defaults to 30. The example below fetches the last week and prints the total cost.
top_models holds up to 10 models, and costs are decimal strings. NagaAI caches the report for up to 5 minutes, so your latest requests can take that long to show up.
To see what one workspace spends, add workspace_id with the workspace’s id from GET /v1/account/workspaces, or default for the keys in no workspace. Every section then counts only the keys that workspace holds now, so a key you move takes its past usage with it. Without workspace_id, the report covers the whole account. Requests from the Playground use no API key and count in the workspace they were made in, which the switcher card picks. See Put keys in workspaces with the API.
Troubleshooting
400 on /v1/account/activity
Thedays value is outside 1 to 30, or workspace_id is something other than a number or default. Send days from 1 to 30, or leave it out to get 30 days. A number that isn’t one of your workspaces isn’t an error: the report comes back with zeros.