Last updated: 1 August 2026
Pare is a browser extension that estimates token usage and cost for AI services you use in Chrome (ChatGPT, Claude.ai, Gemini, and supported API consoles). It runs entirely on your device.
By default, Pare does not transmit any data anywhere. All request metadata Pare captures (timestamps, model names, token counts or estimates, and estimated cost) is stored locally in your browser's IndexedDB storage and never leaves your device unless you explicitly enable sync (see below).
Pare does not capture or store the content of your prompts or responses. It extracts only the metadata needed to estimate cost: which model was called, how many tokens were used (or an estimate, for web apps that don't report token counts), and when the call happened.
If your organisation uses Dendrons' Rein platform, you may optionally enable sync in Pare's settings page. If enabled, Pare sends your usage records to the Rein endpoint and API key you configure - this is your organisation's own infrastructure, not a Dendrons-operated service. You control this setting and can disable it at any time. Sync is off by default and Pare functions fully without it.
Pare fetches a public pricing table (model to cost per token) from a Dendrons-hosted URL to keep cost estimates current. This is a one-way fetch of pricing data; it does not transmit any of your usage data.
Dendrons does not sell, rent, or share your usage data with third parties. We do not have access to your local usage data unless you explicitly enable sync to your own organisation's Rein endpoint.
Since data is stored locally, you control retention - clearing the extension's storage (or uninstalling Pare) removes it. If sync is enabled, retention on the Rein side is governed by your organisation's own policy.
Questions about this policy or Pare's data practices: nivedita@dendrons.ai