WattScope documentation
WattScope estimates the carbon footprint of your organization's AI usage and keeps it attributed to a project and a person, so you can see where it goes rather than only how much there was.
Connecting
WattScope is a remote MCP connector. In Claude, add a custom connector with this URL:
https://www.wattscope.com.au/mcp
Claude will send you here to sign in and to approve the connection. You will be asked to accept the terms once, and again if they change. After that every connection is silent — there is no API key to copy anywhere, and nothing to paste into a configuration file.
Step-by-step for Claude Desktop, Claude Code, Cursor, VS Code and Codex ›
Disconnecting in Claude stops new requests immediately. An access token that has already been issued stays valid until it expires; if you need one cut off sooner than that, tell us and we will do it.
Things to ask for
You do not call these tools yourself — you describe what you want and Claude chooses. These are phrasings that work:
- “Start tracking a new session for the reef survey project.”
- “Record that as 40,000 input and 8,000 output tokens on Sonnet.”
- “What has this session cost in carbon so far?”
- “Show me the whole organization's footprint for this month, split by how well evidenced it is.”
- “Which of our projects is using the most, and how confident are those numbers?”
- “Switch me to the other organization I belong to.”
A session is a piece of work, not a conversation. It lasts until you start another one, so usage recorded across several chats about the same job stays together — and starting a new session is something you ask for, not something that happens when you open a new window.
What the numbers mean
These are estimates, not measurements, and the difference matters enough that every response we return says so.
Providers do not publish per-token energy figures for hosted models, so the per-model values are inferred from public analyses of whole-query energy use. They are uncertain by a factor of several in either direction. Use them for internal awareness and for comparing one period or project against another; do not use them for regulatory reporting, public claims, or offset retirement volumes.
This deployment is running factor set 2026.08.18a-draft,
whose status reads INDICATIVE - order-of-magnitude estimates, not measurements. Suitable for internal awareness and relative comparison; not for regulatory reporting, public claims, or offset retirement volumes.. The same string is returned
to every client that asks, so this page cannot claim more than the service
does.
Why a report shows two figures
Not all usage is evidenced the same way, and adding it together would hide that. Every report separates what is reconciled against a vendor's own invoice from what is not:
| Source | What it is | Billing-grade? |
|---|---|---|
| Vendor Admin API | Token counts pulled from the vendor's own usage API, reconcilable against the bill. The strongest provenance available. | Yes |
| Transcript sync | Token counts read from the conversation transcript the vendor returned. Machine-recorded, but not reconciled against a bill. | No |
| Client-measured (OTLP) | Token counts exported by the AI client's own instrumentation as the work happened. Machine-recorded, and trusted as far as the client is. | No |
| Self-reported | Token counts entered by hand, by a person or by an assistant acting for one. Unverified: nothing checks them against the vendor. | No |
A total that spans both classes is flagged as mixed wherever it appears, including in the CSV export. Only the billing-grade portion is defensible as a measured total — which is why we would rather show you two numbers than one that cannot be quoted.
Your data
- Every figure is scoped to your organization. There is no way for a tool call to see or affect another organization's data, and the test suite treats that as a permanent, measured property rather than an assumption.
- We store what you record: token counts, a model name, a project name, a timestamp. We do not store your prompts or the model's responses. If you upload a claude.ai data export to bring older usage into your history, we do receive it: the file is read into memory to count what it contains, is never written to disk, and is discarded once the counts are stored.
- Each stored figure keeps the version of the factor table that produced it, so revising the table never silently rewrites your history.
- The full statement is in the Privacy Policy, and the Terms say what we commit to.
Exporting
The dashboard exports a CSV of your organization's
usage by project. It carries the billing-grade/indicative split, the factor
set version and the estimate disclaimer in a comment block above the table —
a spreadsheet does not look like an estimate, so the file says what it is.
Readers that skip # lines get a clean rectangle.
Something wrong?
Support and security contact. If a figure looks wrong, tell us what you expected and when you recorded it — the factor set version on the response is the thing that lets us reproduce it.