Usage analytics
Administrators only.
Settings → Usage analytics
Usage analytics gives you an overview of how your team uses 9brains. This lets you track adoption rates, identify popular models and observe usage over time.
Select time range
Section titled “Select time range”At the top edge you select the time range:
- Today
- This month
- Last month
All metrics and charts automatically adjust to the selected time range.
Key metrics (KPIs)
Section titled “Key metrics (KPIs)”Five metric cards give you a quick overview:
| Metric | Description |
|---|---|
| Active users | Number of users who were active in the time range, incl. adoption rate |
| Sessions | Number of started chat sessions |
| Messages | Total number of sent messages |
| Knowledge queries | How often the AI accessed knowledge bases |
| Documents | Total number of documents in knowledge bases, incl. new ones in the time range |
Chats imported from ChatGPT or Claude are not counted here. On import they keep their original dates and would otherwise show usage for periods in which nobody had worked with 9brains yet. This applies to every metric and chart, including model usage. If you continue writing in an imported chat, your new messages count as usual.
Charts and analyses
Section titled “Charts and analyses”Usage trends
Section titled “Usage trends”A line chart shows the usage development over the selected time range. This lets you see at a glance whether usage is rising, staying constant or falling.
Model usage
Section titled “Model usage”A bar chart shows which AI models take up your workspace quota. This helps with the decision of which models should remain activated.
The Cost view is preselected: it shows the share of consumption each model accounts for. That is rarely the same as how often a model is used, because an expensive model carries weight even when it answers rarely. Count switches to the number of answers per model, with each bar also labelled with its percentage share. Both views show shares, never euro amounts.
Tool usage
Section titled “Tool usage”A bar chart shows which integrations and tools are used most.
Knowledge management
Section titled “Knowledge management”A chart shows the trend of knowledge queries, how often the AI falls back on knowledge bases to generate answers.
Quota overview
Section titled “Quota overview”At the top edge of the page you see the total quota of the workspace, the sum of all individual quotas of your licenses. This lets you see at a glance how much of the monthly budget has already been used and how much room remains.
You can find more on the topic of quota and how it is composed under Budget & Quota.
User activity
Section titled “User activity”A detailed table lists the activity per user:
| Column | Description |
|---|---|
| User | Name of the user |
| License | License type (Pro, Business, Max, Knowledge Only) |
| Quota | Individual usage in relation to the quota |
| Sessions | Number of chat sessions |
| Messages | Number of messages |
| Knowledge queries | How often the AI retrieved content from knowledge bases for the user |
| Knowledge contributions | How often the user handed content to the AI for saving into the knowledge base |
| Last activity | When the user was last active |
| Primary model | The most frequently used AI model |
The table is sortable and searchable, so you can search specifically for individual users.
The Quota column is colour-coded: black up to 100%, amber above it, and red from the personal limit onwards (see Budget sharing within the team). Red does not automatically mean blocked. If the workspace has credit, the user keeps working normally, but their usage is then charged to the credit balance instead of their colleagues’ quotas; the cell says so with the note “paid from credit”. Without credit there is a lock icon there, and the user is limited to the Standard model.
Adoption
Section titled “Adoption”Additional metrics on usage adoption help you understand how well 9brains is being adopted by your team and where there is still potential.
Agent overview (admin)
Section titled “Agent overview (admin)”At the bottom of the usage analytics, a table lists all agents in the workspace, regardless of whether they are personal or shared. As an admin, this lets you see at a glance which agents are particularly active, how much they cost and who runs them.
The table shows autonomous runs only: schedule, webhook and runs started via “Run now”, including the sub-agents they spawn. Whatever a person types into a thread counts as their own usage and appears further up under user activity. This also applies when someone keeps writing in a thread an autonomous run created: the autonomous part stays here, the hand-written part goes to that person’s account.
| Column | Meaning |
|---|---|
| Agent | Name and owner of the agent |
| Cost allocation | Owner of the agent, or the workspace for shared agents |
| Usage | Usage of the autonomous runs in the selected time range |
| Cost cap | Monthly safety net (default €10, adjustable at any time by owner or admin) |
| Utilization | Usage in % of the cap; red when close to the limit, yellow when the agent is paused |
| Top model | Which AI model the agent uses most |
| Autonomous runs | Number of autonomous executions in the time range |
| Status | Active or paused |
The table is sortable by name, usage, cap, utilization and number of runs. A totals row at the end shows the overall usage of the autonomous runs and the sum of all cost caps.
Cost cap and utilization refer to the same figure as the usage column. An agent is therefore only slowed down by what it consumes autonomously. The table still lists every agent, including chat-only agents, which then show €0.00 and zero runs.
Admin actions per agent
Section titled “Admin actions per agent”Via the actions menu (three dots) at the end of each row:
- Edit cost cap & models: Admin override independent of the owner. You can lower or raise the monthly cap, set an early warning threshold and restrict the list of allowed models.
- Pause / Resume: Stops the agent immediately or puts it back into operation. For paused agents, neither chats nor executions by schedule or webhook run.
- Delete: Removes the agent permanently. When deleting, triggers are also stopped, webhook URLs invalidated and workspace content (
AGENTS.md, reference files, run results) removed.
These actions act independently of ownership, admins can also override personal agents of other users.
When is this relevant? Typical occasions: an agent consumes more quota than planned; a schedule runs amok; an employee leaves the company and their agents should be cleanly shut down.
Public chats (admin)
Section titled “Public chats (admin)”If your workspace has a public chat, a table of its own below the agents shows what it cost. Without a public chat this section does not appear. Like autonomous agents, public chats are paid from the workspace’s credit, never from your licenses’ quotas, so they count neither towards user activity nor towards the agent overview. During the trial, 9brains instead covers up to 0.50 USD per day for all public chats together.
| Column | Meaning |
|---|---|
| Public chat | Name of the knowledge base published as a public chat |
| Spend / Limit | Spend in the selected period and the monthly limit you set in the public chat’s settings |
| Usage | Spend as % of the limit; yellow from 80 %, red from 100 %. Without a limit this shows a dash |
| Conversations | Number of conversations in the period that incurred costs |
| Model | The model the public chat answers with today; for the standard or premium model, with the current model |
| Status | Published, limit reached or not published |
Status and model always show today’s state, even when you look at last month. “Limit reached” means the public chat is unavailable until the end of the month, or until you raise the limit. A public chat that is no longer published stays in the table as long as it cost something in the selected period. What public chats deleted since have spent is shown in a row of its own. As an administrator, the name takes you straight to the public chat’s settings.
In the quota card further up, the public chats’ spend appears as an amount of its own next to the agents’.