Toktra captures token consumption metadata from every AI tool in your organization — ChatGPT, Claude, Copilot, and more — without ever reading prompt content. You get a real-time picture of who is using what, at what cost, and on which devices, across your entire fleet.Documentation Index
Fetch the complete documentation index at: https://docs.toktra.dev/llms.txt
Use this file to discover all available pages before exploring further.
Key capabilities
Fleet-wide visibility
See LLM usage across all managed devices from a single dashboard. Enrolled devices, active users, and 30-day token totals update in real time.
Shadow AI detection
Toktra reconciles agent telemetry against provider APIs (OpenAI, Anthropic, Azure, Vertex) to surface AI usage on unmanaged devices you didn’t know about.
Usage policies
Define rules that trigger alerts, create incidents, or block requests when usage patterns exceed thresholds. The policy engine evaluates every event as it arrives.
Department budgets
Set monthly token budgets at the org or department level. Soft caps send alerts; hard caps stop requests at the source before they exceed your limit.
Emergency lockout
Revoke a terminated employee’s access to all LLM providers and enrolled devices with a single action. The lockout is logged and auditable.
Privacy-first design
Toktra captures hostnames, byte counts, and token estimates — never prompt content. Employees can pause monitoring during Privacy Hours.
Who Toktra is for
Toktra is designed for IT administrators and security teams at organizations where AI tools are in active use and centralized oversight matters. If you need to answer questions like:- “Which teams are spending the most on LLM APIs?”
- “Is anyone using an AI tool we haven’t approved?”
- “Did this terminated employee’s access get fully revoked?”
Supported platforms
Toktra monitors LLM traffic across all major platforms in your fleet.| Platform | Agent | Coverage |
|---|---|---|
| macOS 13+ | Native agent (Swift + Rust) | System-wide TLS traffic via Network Extension |
| Windows 10+ | WFP driver + Windows Service | Kernel-level outbound network observation |
| Linux (kernel 5.10+) | eBPF daemon | Socket-level tracing via CO-RE eBPF programs |
| Chrome | Browser extension (MV3) | Browser-based LLM requests |
| Edge | Browser extension (MV3) | Browser-based LLM requests |
| Safari | Safari Web Extension | Browser-based LLM requests |
| Firefox | WebExtension | Browser-based LLM requests |
All platform agents share a common Rust core library for event buffering, token estimation, protobuf serialization, and mutual TLS authentication. Telemetry from every platform flows into the same dashboard.
Next steps
Quickstart
See your first usage event in under 10 minutes.
Deploy agents
Roll out agents across your fleet using MDM, Intune, or your package manager.
How it works
Understand the data flow from device to dashboard.
Privacy model
Learn what Toktra captures and what it never touches.