opencode-usage-panel

A sidebar panel for the opencode TUI that breaks a session down by model: how many tokens each one consumed and what it cost.

The Context block mirrors opencode's built-in panel — the token count and share of the context window for the latest assistant turn — but leaves out its cost, which does not account for peak pricing.

Each model gets a header row with its estimated cost, then the input (cache miss), output, reasoning, and cache read (cache hit) counters summed over the session. The percentage on each row is that counter's share of the model's own total tokens.

Requirements

• opencode >= 1.18.0

The plugin is a TUI plugin: it imports from @opencode-ai/plugin/tui and renders into the sidebar_content slot at order 100, the same slot the built-in context panel uses.

Install

From npm

Available on npm: opencode-usage-panel.

Install it with the CLI:

opencode plugin opencode-usage-panel --global

Or add the package to tui.json yourself — opencode installs npm plugins on startup. Use .opencode/tui.json (project) or ~/.config/opencode/tui.json (global):

{
  "$schema": "https://opencode.ai/tui.json",
  "plugin": ["opencode-usage-panel"]
}

From a local checkout

There's no need to copy the file into a plugin directory. Point tui.json straight at the source with an absolute path (or a file:// URL):

{
  "$schema": "https://opencode.ai/tui.json",
  "plugin": ["/Users/you/opencode-usage-panel/src/index.tsx"]
}

Disable the built-in panel

The plugin renders into the same slot as opencode's built-in context panel, so turn that one off to avoid stacking:

{
  "plugin_enabled": {
    "internal:sidebar-context": false
  }
}

Pricing

Cost is computed per message from a rate table, so peak/off-peak pricing is honoured using each message's own timestamp. Models without a configured rate table fall back to the cost opencode records, which is priced automatically from models.dev (free tiers show —).

Rate tables

Prices are US dollars per million tokens. input is cache-miss input, cacheRead is cache-hit input, and output also covers reasoning unless a separate reasoning rate is given.

{
  "plugin": [
    [
      "opencode-usage-panel",
      {
        "pricing": {
          "deepseek-flash": {
            "input": 0.15,
            "output": 0.6,
            "cacheRead": 0.003,
            "peak": {
              "input": 0.3,
              "output": 1.2,
              "cacheRead": 0.006
            }
          }
        }
      }
    ]
  ]
}

Keys can be a bare modelID or providerID/modelID. Top-level rates apply outside peak windows; the optional peak block replaces them inside peak windows. A model without peak has flat pricing.

Peak windows

peakHours is always specified in UTC, since that is how providers publish their schedules. Windows may wrap past midnight, and days restricts them to UTC weekdays (mon..sun, JS-style 0..6 also accepted; omit for every day). DeepSeek's schedule — weekdays 01:00-04:00 and 06:00-10:00 UTC — is:

{
  "peakHours": [
    { "start": "01:00", "end": "04:00", "days": ["mon", "tue", "wed", "thu", "fri"] },
    { "start": "06:00", "end": "10:00", "days": ["mon", "tue", "wed", "thu", "fri"] }
  ]
}

Messages are matched against these windows in UTC, so the result is correct whatever your machine's timezone is. For display, times are converted to your local timezone by default (IST for Asia/Kolkata); set "timezone": "utc" to render them in UTC instead.

The extra cost paid for peak messages is shown under Spent. While a peak window is active the panel shows a ▲ peak badge next to affected models and the time the window ends; otherwise it lists the configured peak windows.

License

MIT — see LICENSE.