Qwen Code

UsageNow shows your Qwen Code activity for today — tokens, answers, the models that gave them, and what that work would have cost at API prices — and a 30-day history. Everything comes from the session recordings Qwen Code writes on your Mac. It doesn't show a usage limit.

Requirements#

• Qwen Code installed — the qwen CLI
• Used at least once, so it has recorded a session

Today's activity#

Qwen Code records every message of a session, and attaches the model's token counts to each answer. UsageNow adds them up for:

• tokens today — the whole prompt, cached part included, plus output and thinking
• requests — answers recorded today
• most recent model and activity by model
• ≈ cost at API prices — see Estimated cost

A session you resume repeats its earlier records; UsageNow counts each answer once. A model UsageNow has no published price for is left out of the estimate, which then reads partial.

What is read, and what isn't#

From each line of a recording, UsageNow decodes only the answer's identifier, its time, the model, and the token counts. The message itself — your prompt, the reply, tool output — is never decoded. Workflow journals kept in the same folders aren't read at all.

Qwen Code's sign-in (~/.qwen/oauth_creds.json) and its settings are never opened.

No usage limit#

Qwen Code doesn't write a limit to disk, so the Qwen Code section says Usage limits unavailable. Today's tokens, requests, and models still appear.

Where UsageNow looks#

UsageNow reads ~/.qwen/projects/<project>/chats/<session>.jsonl. It recognizes Qwen Code as installed when ~/.qwen exists, or when the qwen CLI is in ~/.local/bin, Homebrew, /usr/local/bin, Bun, or a Node version under nvm.

States#