What Kenaz reports to your organization’s Kameas Fleet
Two Kameas programs can send usage telemetry to your organization’s Kameas Fleet: the Kenaz Harness, and the kenaz program, which connects AI tools made by other companies to Kameas Fleet. Both send only the fixed set of records below, and only for the classes each person turns on. Every class is off until you turn it on.
The kenaz program reports only on the AI tools that our Supported Tools page marks as reporting usage, and for each of them it sends only the records below. Adding a tool to that page does not add anything to this list. Estimated cost is passed on only when the tool reports it; the kenaz program does not estimate prices.
| Class | What is sent | Plans |
|---|---|---|
| Usage counts | session started/ended; duration; input and output token counts; estimated cost; whether the cost is known; model provider | Pro and above |
| Tool calls | tool kind from a fixed list; success; duration | Pro and above |
| Errors | error category from a fixed list; whether recoverable | Pro and above |
| Model calls | model name and provider (and, for a router, the provider it chose); request settings (maximum tokens, temperature); token counts (input, output, cached, reasoning); number of tools offered and an estimate of their size; estimated cost; time to first token; duration; finish reason; error category; whether the call was streamed, retried or a fallback; the provider’s system fingerprint; a keyed hash of the session, which groups calls from one session without revealing the session ID | Pro and above |
| Diagnostics (Kenaz Harness only) | task, hook and command timings and outcomes | Team and above |
Every record also carries: the person’s account ID, the organization, an opaque device ID, the program that produced it (kenaz-harness for the Kenaz Harness, or kameas-fleet for the kenaz program) and its version, which AI tool the record is about and that tool’s version, and inside a Kenaz Workbench the workbench ID. When a device connects it also sends a device display name, the operating system and the architecture.
Never sent: prompts; AI responses; tool inputs or outputs; file contents; source code; file paths, folders or repository names; error message text; names of tools or servers you added; identifiers a model provider assigns to responses.
How we make sure: filtering happens on your device, in open-source code: kenaz-harness and kenaz-plugin, internal/project. Our test suite plants marker strings in prompts, tool output, paths and errors, and fails if any of them reaches Fleet. Fleet then checks every record against a closed schema and rejects anything else. Fleet also maps any tool name outside its fixed list to “mcp” or “other” on arrival.
What the kenaz program reports about your sessions is decided by open-source code you can read at github.com/kameas-ai/kenaz-plugin. The part that connects to Kameas Fleet (sign-in and transport) is proprietary.
Who sees it, and how long it is kept: see your organization’s agreement and your notice in the app. In short: administrators see team-level totals. Data is kept while your organization subscribes, and deleted within 72 hours when you opt out of a class or leave.
Not telemetry, and not covered here: Offload (workflow data for suggestions) is a separate, organization-enabled feature with its own notice. Memory sync, which copies the memory notes you choose to your own Fleet account, is off until you turn it on and has its own notice. Skills you publish and team context you share are sent because you choose to send them. Hosted connectors, which your organization can choose to turn on, are described in our Privacy Policy (§13) and on the Supported Tools page.
Kenaz and Kameas are trademarks of Kameas AI, Inc. Names of AI tools and platforms made by other companies are trademarks of their respective owners and are used only to identify those products. Kenaz is an independent product of Kameas AI, Inc. and is not affiliated with, endorsed or sponsored by any of them.