Incident Management
Platform
AI
Dedicated compute
Dedicated compute keeps concurrent queries stable and complex analysis fast.
Learn moreDedicated compute
Dedicated compute keeps concurrent queries stable and complex analysis fast.
Learn moreAI usage and trace collection client
Start on each user’s computer. Bring Tokens, costs and traces from different AI tools into one workspace, with usage attributed to the people behind it.
Capabilities
Accounts are provisioned, quotas are approved and the bill arrives every month. What matters next: is AI moving tasks forward, or repeatedly processing the same context?
Your everyday AI tools, in one view
Codex, Claude Code, Kiro, WorkBuddy and Qoder, plus plugins, SDKs and APIs connected through their supported integrations.
Break down Tokens, requests and costs by member, tool and model. See who is using resources and where that usage is going.
Recorded usage from one intensive user
A usage calendar reveals consumption spikes
223M input / 864K output
Drill down from totals to individual requests
Follow the journey from request to response. Inspect P50 and P75 latency, then drill down to model and tool calls.
Separate frequently used models from the long tail. Inspect wait and sleep calls to uncover time spent waiting or repeating operations.
Track consumption and remaining quota across 5-hour and 7-day windows. Configure alerts before limits interrupt the work.
Quota windows and reset-credit information depend on support from the connected tool and account.
Quota windows
View remaining reset credits alongside quota information, so teams can plan resource use before an interruption.
A vendor’s dashboard explains usage within its own product. AgentMeter collects data from connected tools at the user’s computer, associating course accounts, team accounts and individually used tools with members for cross-product analysis.
Use the desktop client or Terminal for individual installation or team rollout. Configuration can associate data with a member UUID and username.
Accounts, plugins, SDKs and APIs
Local collection · JSON configuration
Member attribution · Shared analysis
AI teaching in practice
A university provided AI tool accounts to more than 100 students. Halfway through the term, teachers still lacked evidence for quota allocation, whether heavy usage meant productive work, and which students needed help.
Some students repeatedly submitted full project context, with input Tokens exceeding output Tokens by more than 200 times.
Some students’ 7-day quota usage stayed in single digits after an unsuccessful first experience.
After rollout, teaching assistants coached heavy users on task decomposition and context management, offered a tool-skills class to less active students, and used usage data to adjust quotas.
Assignment scores in this group improved. Usage totals, active days, daily peaks and tool distribution also informed the next term’s course design.
AgentMeter · v0.1.8
Install AgentMeter on the computer running your AI tools, then import the dedicated JSON configuration from your workspace.
For Apple Silicon and Intel Macs
For x64 and ARM64 devices
For Debian / Ubuntu, RHEL / Fedora and related distributions
Choose the package that matches your processor architecture. Linux packages are available in DEB and RPM formats.
After installation, get your configuration in Agent MonitoringIn the Guance console, open Agent Monitoring. Choose Connect Agent or the “+” on an Agent application, then select the Agent to connect.
Download the client for your operating system and processor architecture, then import the dedicated JSON configuration. Terminal setup is also available through the console guide.
Optionally associate a member UUID and username. View usage, request traces, quota trends and tool distribution in your workspace.
