Sam LLM Assistant speaks the Anthropic Messages API and routes your traffic to NVIDIA NIM, OpenRouter, DeepSeek, Ollama and more - with API keys, authentication and usage analytics built in.
$100.00 in free credits on signup
# Point the Anthropic SDK at your proxy - nothing else changes
curl https://api-proxy/v1/messages \
-H "x-api-key: sm_oa_•••" \
-H "anthropic-version: 2026-06-01" \
-d '{ "model": "claude-3-5-sonnet", "max_tokens": 512,
"messages": [{"role":"user","content":"Hello!"}] }'Drop-in compatible, provider-agnostic, and observable from day one.
Drop-in /v1/messages endpoint. Keep your Claude SDK and tooling - just change the base URL.
Route to NVIDIA NIM, OpenRouter, DeepSeek, Ollama and more. Switch providers without touching app code.
Native server-sent events, so tokens arrive the moment the upstream emits them. No buffering.
Issue, name, disable and expire keys per user. Only salted SHA-256 hashes are ever stored.
Every request is logged - tokens in/out, model, provider, status - so you can see exactly what costs what.
Run the proxy anywhere and host the PHP dashboard on your own stack. Your keys never leave your infra.
No SDK rewrites. Keep your Anthropic client, change one base URL.
Register in seconds.
Mint an sm_oa key from your dashboard. Copy it once - only the hash is stored.
Set your Anthropic client base URL to the proxy and pass the key. That is the whole integration.
Route to whoever is fastest or cheapest today - switch any time without touching your app code.
Verify your Telegram to unlock your signup credits, then share your invite link to keep earning.
Create an account and verify your Telegram - your balance is credited instantly, ready to spend on the API.
Share your invite link. When a friend signs up and verifies their Telegram, you both get paid - instantly.
Create an account, generate a key, and point your Anthropic SDK at the proxy.
No. Sam LLM Assistant implements the Anthropic Messages API, so you keep your existing Claude SDK or HTTP client and only change the base URL and key.
NVIDIA NIM, OpenRouter, DeepSeek, Ollama and any OpenAI-compatible endpoint. You can switch providers from configuration without redeploying your app.
Raw keys are shown once at creation. The server only ever stores a SHA-256 hash, so a database leak never exposes a usable key.
Yes - responses stream over server-sent events end to end, so your users see tokens as they are generated.
When you verify your Telegram, your account is credited with free signup credits. Every API request spends from that balance. When it reaches $0.00 the key is paused until your balance is topped up again or refreshed.