Skip to content

Requesty

Requesty is a router. One key reaches many models across many vendors, and Requesty picks the upstream.

Requesty is a passthrough provider. Its router speaks the native Anthropic Messages protocol, so Ptah sends requests straight through with no translation layer in between. That makes it one of the lowest-overhead multi-model paths in Ptah — unlike OpenRouter and Sakana, which need a local translation proxy.

An API key from app.requesty.ai/api-keys.

  1. Open Settings → Providers → Requesty.
  2. Paste your API key.
  3. Click Test connection.

Or from the CLI:

Terminal window
ptah provider set-key --provider requesty --key <your-key>
ptah provider default set requesty

The key is stored in the operating system’s secure credential store, never in a settings file.

Requesty has no static model list in Ptah. The catalog comes live from Requesty’s /v1/models endpoint, so the picker always shows what your account can actually reach.

Ptah maps the tier words opus, sonnet, and haiku onto real model ids.

Requesty ships no tier map, deliberately. The only model slug in Requesty’s documentation appears on its OpenAI-compatible lane, not on the Anthropic passthrough lane that Ptah uses. Shipping an unverified id as a default would silently break every “Default (recommended)” selection.

Instead, Ptah derives the mapping from your live catalog. If that derivation picks wrong, set the tiers explicitly from the model picker:

Terminal window
ptah provider tier set --tier opus --model <model-id>

An explicit choice always outranks a derived one.

Requesty bills you directly. Ptah reports what the router reports.

A 404 on the first message — the tier word did not resolve to a servable model id. Pick a model explicitly rather than leaving it on “Default”, or set the tier map as shown above.

“Unauthorized” — the key is wrong or revoked. Regenerate it at app.requesty.ai/api-keys and re-paste it.