ProxonProxon

The more Proxon observes, the more it can tell you.

Proxon collects AI usage from several kinds of sources. Some watch where your people work, others pull from your vendors or your own agents. The more you connect, the stronger the signal: broader coverage of how AI is being used, and sharper detail on what it costs. You don’t need everything to begin — and you don’t need all of it ever. Pick the sources that fit your organization and add more when you want more coverage.

You stay in control · for IT & Security

Every observer can be turned on or off, and one capture level, set by an admin, applies to all three. Each level includes everything in the level before it. Content aware is the default, and an admin can lower it at any time.

  • Metadata only. The facts of an interaction: which AI tool, when, how often, and how much (token counts and cost), without any of the words. It’s what drives adoption and most cost signals.

  • Behavioral data. Adds the shape of each interaction: how big a prompt, file, or paste was, which tools ran and for how long, and which provider it went to. Still none of the words themselves.

  • Content aware. Adds the words: prompts, replies, the model’s reasoning, and what tools were given and returned, plus document and tab titles and full page addresses. This is the default level, and an admin can lower it at any time. What each observer adds at this level is spelled out on its card below.

No observer records which keys you press. The observers capture the words you type into AI tools, like your prompts, only at the Content aware level, and only on supported AI sites and apps. At every level, the Network Observer strips secrets such as API keys, passwords, and tokens out of captured text before it leaves the device.

Admins manage all of this in Data Collection under Org Settings and can change the level at any time.

See exactly what each level captures →

Recommended

These cover most organizations, and most rollouts start here. The three observers each watch a different place where your people use AI (the browser, the desktop, and the network), and each catches usage the other two can't see. AI vendor integrations add usage and spend straight from the vendors you already pay, with nothing installed on anyone's device.

Browser Observer

extension for Chrome, Edge, Firefox & Safari
Uniquely collects
AI used inside the browser: which web assistants (ChatGPT, Claude, Gemini, and others) people are on, plus engagement signals like which tab is active and how actively it is being used. The only source that sees browser-based AI at all.
How it works
Runs as a browser extension, observing AI tab activity locally in the browser. Below the Content aware level it records a site’s domain, not the full page address or the tab title. At the Content aware level it also keeps the full address and tab title, and on supported AI chat sites, the prompt text.

Desktop Observer

lightweight agent, macOS & Windows
Uniquely collects
Activity and context across every desktop app: which app is in the foreground (and, at the Content aware level, that window’s title), plus how actively the device is being used. Reveals workflow patterns like moving an answer from an AI app into a document. The only source that captures what’s happening around AI usage, including in non-AI apps.
How it works
A lightweight agent that runs on the device and observes which app is in the foreground plus engagement signals. At the Content aware level it also records the foreground window’s title, which can include a document’s name.

Network Observer

on-device capture for CLI & desktop apps
Uniquely collects
The content of AI interactions (prompt text, model responses, and tool-use calls), plus token counts for supported apps. Those token counts make granular, per-request cost attribution possible for CLI and desktop AI (supported apps only; some don’t expose usage).
How it works
Inspects calls to a defined allowlist of known AI hosts at the capture level your admins set; everything off the list passes through untouched: for that traffic Proxon sees the connection exists, not its contents. Apps that pin their certificates can’t be inspected. Supports macOS and Windows.

AI vendor integrations

read-only links to the AI vendors you pay
Uniquely collects
Usage and spend straight from the vendor’s own admin and billing data, the most accurate cost source for the accounts you connect. Captures usage that never runs on an employee’s machine, like server-side or automated calls.
How it works
An admin connects each vendor once from the Integrations page, usually with a read-only key created in the vendor’s own admin console. Proxon pulls from the vendor directly, and nothing installs on anyone’s device.
Set it up →

Options for developers

These cover the AI your engineering teams run themselves. Use the Capture Proxy or OTel for AI Tools to bring AI coding tools like Claude Code and Codex into Proxon, and the Proxon SDK for Agents or OTel for Agents for the agents and services you build in-house. Those often run server-side or in CI, where the observers can't see them. Each carries only what it is given: the SDK never sends prompt or reply content, and OpenTelemetry sends only what the tool or exporter includes.

Capture Proxy

change one base URL in your AI coding tool
Uniquely collects
Prompts, replies, and cost from AI coding tools like Claude Code and Codex CLI, and from your own SDK apps, with nothing installed on the machine.
How it works
Each developer gets a personal URL from Preferences and sets it as the tool’s base URL. Requests pass through to the provider unchanged, and signing in works the same as before.
Set it up →

Proxon SDK for Agents

a drop-in SDK for your own agents
Uniquely collects
Per-agent, per-task economics from inside the agents you build: exact cost and token usage tied to a specific agent and task, never prompt or response content. Sees autonomous, programmatic activity the device observers (which watch people) can’t.
How it works
A Node / TypeScript SDK: add one launch flag and it auto-instruments your Anthropic and OpenAI clients with no call-site changes, then reports directly to Proxon. Runs anywhere Node does: CI, containers, servers, or a laptop. For other languages, use OTel for Agents.
Set it up →

OTel for Agents

point your existing OTLP exporter at Proxon
Uniquely collects
Model usage and cost from the agents and services you build, in any language, including server-side AI that never runs on anyone’s device, like a backend service, a CI job, or a scheduled agent. You control what your exporter sends; Proxon only needs the model and token counts.
How it works
An admin mints an organization credential, and you set a few standard OTLP environment variables on your existing exporter: no new agent, no proxy, no certificate. Usage sent this way belongs to the organization, not to any one person.

The Proxon SDK and OTel for Agents do the same job two ways. For a Node app you control, the SDK auto-instruments it with one flag. For anything else (another language, or a service that already emits OpenTelemetry), use OTel for Agents. Either is enough; running both adds detail, not new coverage.

Set it up →

OTel for AI Tools

Claude Code & Codex report their own usage
Uniquely collects
Usage and cost straight from Claude Code and Codex, in the terminal and in their desktop apps, credited to the person who set it up. A way to see these tools without putting an observer on the machine.
How it works
Each developer creates a personal credential in Preferences and runs one command that adds a few settings to the tool’s own config. There is no app to install, nothing running in the background, and no administrator in the path.
Set it up →
Where to start

Most teams start with the recommended sources: the observers install through the tools you already use, vendor integrations install nothing, and together they give broad coverage on their own. Add an option for developers when you want to see AI coding tools, your own agents, or backend services. You’re never locked in: admins control the observers and their capture level at any time.