Metered inference
The models you already use. One key.
Keep the client you have. One base URL reaches every model in the catalog, every request runs under a zero-retention agreement, and the prices are on this page before you sign up.
- OpenAI and Anthropic compatible
- Zero retention
- Prices per 1M tokens, in euro
02 / Ledger
Price per 1M tokens
Every model the API serves, priced per million tokens. Cached input is the rate when the provider serves a prefix from its cache.
Updated
- 1M context
- 131.1K max output
- Tools
- Reasoning
- 1M context
- 131.1K max output
- Tools
- Reasoning
- 1M context
- 393.2K max output
- Vision
- Tools
- Reasoning
- 1M context
- 393.2K max output
- Vision
- Tools
- Reasoning
- 1M context
- 131.1K max output
- Vision
- Tools
- Reasoning
- 1M context
- 131.1K max output
- Vision
- Tools
- Reasoning
- 1M context
- 131.1K max output
- Tools
- Reasoning
Every response reports what it cost, in euro.
03 / Integrations
Point your tools at zdr.sh
OpenAI compatible at https://api.zdr.sh/v1 and Anthropic compatible at https://api.zdr.sh. Streaming and tool calls pass through unchanged, and each guide has the full setup.
04 / Notes
Three things worth knowing
Nothing kept, by design
Every provider is under a zero-retention agreement. We do not collect prompts or answers, and there is no trace system to opt out of.
A memory you can open
A notes workspace your models can read and write. Search it, write to it, and see how the notes connect; it belongs to the account, not to a vendor.
The tools you already use
OpenAI and Anthropic compatible, so opencode, Claude Code, Hermes Agent, the SDKs and cURL work by pointing their base URL at the API.
05 / Start
One key, then a request.
Create a key, point a client at the base URL, and the quickstart walks the first request through. It takes about a minute.