Watch your AI subscription usage
from anywhere
隨時隨地
掌握你的 AI 訂閱額度
Ravenclaude publishes only the data you choose* to a cloud server, so you can see it across every machine and on your phone. Ravenclaude 只傳送你選擇的資料*到雲端伺服器,讓你在每一台機器、甚至手機上都看得到。
*By default only your session id and rate-limit windows (with reset times) are sent. Cost and context are opt-in. *預設只會送出 session id 與 rate-limit 視窗(含重置時間);花費與 context 由你自行開啟。
Request access申請內測資格
Enter your email and we'll send a one-time link to claim your access token. Seats are limited. 留下你的 email,我們會寄出一次性連結,開啟後即可領取 token。名額有限。
- This is an alpha test. Data is not guaranteed to be retained and may be wiped at any time. 這是 alpha 測試,資料不保證保留,可能隨時被清除。
- Service quality and availability are not guaranteed; the API may change or go down without notice. 不保證服務品質與可用性,API 可能隨時變動或中斷,恕不另行通知。
- Your email is used only to send the claim link and test-related notices, and is deleted when the test ends, except limited records we must briefly keep for security or legal reasons. 你的 email 只用於寄送領取連結與測試相關通知,並於測試結束後刪除;因安全或法律需要而必須短暫保留的有限紀錄除外。
Quick setup — about 5 minutes5 分鐘快速設定
1Claim your token領取你的 token
Open the link in the email and click Reveal my token. It shows once — copy it somewhere safe. This is your PUBLISHER_TOKEN.
開啟信中的連結,點 Reveal my token。token 只會顯示一次,請立即複製保存,這就是你的 PUBLISHER_TOKEN。
One token = one subscription account. Don't publish Claude and Codex through the same token — their live rate-limit windows would overwrite each other. During the alpha each account gets a single token, so pick one account to connect for now. 一個 token 只對應一個訂閱帳號。請勿用同一個 token 同時上傳 Claude 與 Codex——兩者的即時 rate-limit 視窗會互相覆蓋。內測期間每個帳號只會有一個 token,因此請先連接一個帳號就好。
2Install the collector (steward)安裝收集器(steward)
Caches your AI usage data locally. Installs a prebuilt program. 在本機暫存 AI 資料。安裝編譯好的程式。
curl -fsSL https://raw.githubusercontent.com/Ravenclaude/steward/main/install.sh | bashThe installer asks: Claude, Codex, or both.安裝時會問:Claude、Codex,或兩者都裝。
3Install the publisher (ranger)安裝上傳器(ranger)
Uploads the cache to the alpha server, in the background — it never blocks your agent. 在背景上傳快取資料,完全不會卡住你的 agent。
curl -fsSL https://raw.githubusercontent.com/Ravenclaude/ranger/main/install.sh | bash4Point it at the alpha server設定連到 alpha 伺服器
Edit ~/.claude/claude-ranger.conf (for Codex: ~/.codex/codex-ranger.conf) and set these two lines — paste the token from step 1:
編輯 ~/.claude/claude-ranger.conf(Codex 為 ~/.codex/codex-ranger.conf),填入這兩行,token 用步驟 1 領到的:
PUBLISHER_URL="https://castle-alpha.ravenclaude.com"
PUBLISHER_TOKEN="paste-your-token-here"Then restart your agent. By default only session_id and rate_limits are sent — never your code, paths, or prompts.
然後重啟你的 agent。預設只會送出 session_id 與 rate_limits,絕不會傳送你的程式碼、路徑或提示。
5Confirm it's flowing確認資料有進來
Use your agent for a moment, then check the server sees you: 繼續用你的 AI agent,然後測試伺服器是否收到資料:
curl -H "Authorization: Bearer <your-token>" \
https://castle-alpha.ravenclaude.com/latestSeeing your rate-limit windows means the pipeline works. To watch it live, install the Ravenclaude app (iOS, with macOS / Android / Windows ports). Point it at the same server URL and token. 看到 rate-limit 就代表資料已經正常上傳,可以用 app 看資料了。要即時查看,請安裝 Ravenclaude app(iOS,另有 macOS / Android / Windows 版本),填入相同的伺服器 URL 與 token 即可。