Get Started with Token Usage Analytics
Token Usage Analytics lives on the Tokens page under Monitor. It is on for every organization; there is nothing to enable. Admins use that page to analyze token usage across MCP servers, toolkits, skills, and tool responses.
Prerequisites
You need:
- admin access to Willow
- an organization with MCP servers, toolkits, skills, or tool calls to analyze
- recent activity in your organization, so the usage tables and recommendations have data to show
If your organization has not made any tool calls yet, or if Willow has not yet counted tokens for any session, the analytics pages have no data to show. They fill in as soon as Willow records new activity.
Open the Tokens page
Open Monitor > Tokens. Use the page to review recommendations and to examine each token source in detail.
What the page shows
The Tokens page contains these areas. Some stay empty until Willow has the data for them:
| Area | Purpose |
|---|---|
| Dashboard | Your estimated monthly spend, the savings available, the three cost drivers in order of cost, and a summary of context health. |
| Analytics | Measured usage across three tabs: MCP Gateway, Conversations, and Devices. |
| Recommended actions | Changes you can make, ordered by how much they would save each month. These appear on the Optimize view. |
| Usage policies | Two limits: a maximum length for conversations, and a monthly budget for each model. The Usage Hooks plugin applies both. See Usage Policies. |
| Context inventory | MCP servers, toolkits, skills, and tool responses ranked by token footprint. |
| Output format | The format Willow currently uses for tool output. Changing this format can reduce the tokens in tool responses. |
| Devices | AI agent token usage collected from managed devices. This area needs AI Discovery scan agents to be deployed and reporting. If you use only the gateway and have no scan agents, this area stays empty. |