OpenCode Go Stats

Adds token and cost breakdowns to opencode.ai workspace usage pages.

There are three ways to run it:

Version	Best for	Output
Browser extension	Keeping stats on the usage page	Inline tables and charts
Userscript	Getting the same dashboard through a userscript manager	Inline tables and charts
Console script	One-off inspection from DevTools	console.table() reports and session totals

The extension and userscript share the same dashboard code. The console script is separate: it prints a report, exposes raw data on window.__opencodeStats, and includes the session breakdown.

Tables

The dashboard adds two tables above the normal usage list.

Table	What it shows
Estimated Pricing	Per-model input, output, and cache-read prices in dollars per 1M tokens. These are fitted from usage records that include cost data.
Per-Model Summary	Requests, input tokens, output tokens, reasoning tokens, cache reads, total cost, and effective dollars per 1M tokens for each model.

The summary table also includes the fitted pricing columns when pricing can be estimated for the model.

Charts

The dashboard has range filters for All, Today, 7d, 30d, 90d, and 1y. The selected range updates the tables and the chart together.

Three of the chart modes are daily stacked bars, split by model:

Chart	What it answers
Cost	How much each model cost per day.
Tokens	How many input, output, reasoning, and cache-read tokens each model used per day, summed into one token total.
Requests	How many requests each model handled per day.

Two chart modes compare models across the selected range:

Chart	What it answers
Efficiency	Which models were most expensive per 1M tokens. This is a horizontal bar chart sorted by effective dollars per 1M tokens.
Share	Which models accounted for the largest share of total cost. This is a horizontal bar chart with percent share and dollar cost in the tooltip.

Example chart views:

Cost	Tokens	Efficiency

Console Script

dist/pull-stats.js is for quick checks without installing anything. Open an opencode.ai workspace usage page, paste the script into the browser console, and run it.

It prints:

Section	What it contains
Estimated Pricing	The same fitted per-model prices used by the dashboard.
Per-Model Summary	Request, token, cost, and effective price totals by model.
Grand Total	Workspace-wide token and cost totals.
Top Sessions by Cost	The 20 most expensive model/session pairs. This is console-only.

After it finishes, the parsed records and computed summaries are available as window.__opencodeStats.

Install

Prerequisite: Build the extension/userscript first with npm install && npm run build.

Browser Extension

Load the dist/extension/ directory as an unpacked extension.

Browser	Steps
Chrome	Open Extensions, enable Developer mode, choose Load unpacked, then select dist/extension/.
Firefox	Open about:debugging, choose This Firefox, choose Load Temporary Add-on, then select dist/extension/manifest.json.

The extension runs on https://opencode.ai/workspace/*/usage.

Userscript

Install dist/opencode-stats.user.js with Tampermonkey, Violentmonkey, or another userscript manager.

The userscript runs on the same usage pages as the extension and renders the same dashboard.

Console Script

Open an opencode.ai workspace usage page, paste the contents of dist/pull-stats.js into DevTools, and run it.

Build

npm install
npm run build

npm run build produces:

Output	Source entry	Purpose
dist/extension/content.js	src/entries/extension.ts	Browser extension content script.
dist/opencode-stats.user.js	src/entries/extension.ts	Userscript dashboard.
dist/pull-stats.js	src/entries/console.ts	Console-only report.

All output goes to dist/.

Release

npm run release

This builds the extension and writes extension.zip.

Type Check

npm run check

API Authentication

The FN_ID constant (see src/packages/core/constants.ts) is the server function identifier used by opencode.ai's internal API gateway to route usage-history requests. It is a public routing key, not a secret, and is embedded in the opencode.ai web client. If the upstream API changes, this value must be updated in src/packages/core/constants.ts.

How It Works

1. Reads the workspace ID from the current usage page URL.
2. Fetches usage pages through opencode.ai's internal /_server endpoint.
3. Deduplicates records and merges them with records cached in localStorage.
4. Computes per-model request, token, and cost totals.
5. Estimates per-model token prices by solving a linear system from records with cost data.
6. Renders the dashboard or prints the console report, depending on the script being used.

Pricing is estimated from your recorded usage, so a model may not show fitted prices until there is enough cost-bearing data for that model.