0x24 watches your Anthropic, OpenAI and Gemini usage through their read-only reporting APIs and alerts you the hour something goes wrong: a runaway agent loop, a leaked key, a misconfigured batch job.
Read-only by construction · keys stored encrypted · no agents, no SDKs, no code changes
0x24 is the character code of $ — watching the dollar sign is the whole job.
No SDK, no proxy, no code changes. 0x24 reads the same reporting APIs your provider's billing console uses.
Create a dedicated admin key with reporting-only scope in your provider's console and paste it once. It's verified live and stored encrypted.
Every few minutes 0x24 pulls cost reports (billing truth) and hourly usage (fast signal) and builds your 7-day baseline per provider.
Spikes vs baseline and budget thresholds (50/80/100%) land in your dashboard — and in Telegram, within minutes of the anomaly.
The invoice at month-end is an autopsy. 0x24 is the smoke detector.
Only reporting endpoints, ever. 0x24 can't touch your workloads, keys or billing — the code has no write path to your providers.
Current hour vs your trailing 7-day baseline. New accounts stay silent until a baseline exists — no alert spam on day one.
Daily and monthly budgets with 50/80/100% thresholds, each fired once per period. Powered by provider cost reports — the billing truth.
Provider keys are encrypted at rest and never displayed again — only a 4-character hint. Use reporting-scope keys, revocable any time.
Link Telegram in one click and alerts reach your phone in seconds. Dashboard keeps the full history either way.
Anthropic, OpenAI and Google (Gemini) today; OpenRouter and Bedrock on the roadmap. One baseline, one alert channel, all your LLM spend.