The team has used
of the 5-hour window. Resets in —.
Last 24 hours
24-hour usage
Daily activity
By user
By model
Activity
Endpoints
Both forwards accept the Anthropic Messages API at
/v1/messages. They differ only in how much of the calling
client's identity reaches the upstream.
/iponly/*Rewrites the source IP only. User-Agent, client fingerprint and body metadata pass through verbatim, and the requested model is never altered. Use this from a real Claude Code install.
/restrict/*Applies one canonical Claude Code fingerprint to every request, and —
for Claude Code callers — the model rewrites listed below. Also hosts the
Codex passthrough at /restrict/backend-api/* and image
generation at /restrict/v1/images/generations.
Models
Matched by prefix — any version under these families is accepted and served by the account's Max subscription. Because they draw on that subscription, they are reserved for Claude Code (CLI or desktop); any other client requesting one gets a 400 and should pick a third-party model instead.
| Model | Served by | Notes |
|---|
On /restrict, requests from Claude Code have
these model ids substituted before dispatch. Sources without a real Anthropic
model behind them exist purely as aliases for reaching third-party upstreams.
Every other client is forwarded the model it literally asked for.
| Requested | Actually runs | Served by |
|---|
The size parameter also accepts the aliases
landscape, portrait and square.
Calling it
Point any Anthropic-compatible client at the proxy base URL.
curl https://claude.cognitiolabs.ai/iponly/v1/messages \
-H "content-type: application/json" \
-H "anthropic-version: 2023-06-01" \
-d '{
"model": "claude-sonnet-4-6",
"max_tokens": 256,
"messages": [{"role": "user", "content": "Hello"}]
}'An OpenAI-compatible facade is available at
/v1/chat/completions, translating to whichever upstream serves
the requested model.