ZeroCredit AIzerocredit.ai
Start free

Migration

Bring your keys. Point your API. Optimize spend.

ZeroCredit works two ways: connect your provider API keys inside the dashboard, or point your existing client at one base URL. Either path speaks the standard chat API, so every request — whichever model it targets — keeps billing through your own provider accounts while gaining semantic caching, budget controls, fallback chains, intelligent routing, ML-powered cost optimization and per-request savings proof. Nothing else in your codebase changes.

OpenAI (GPT)Anthropic (Claude)Google (Gemini)xAI (Grok)DeepSeekKimiPerplexityAzure OpenAI
  1. 1

    Create a ZeroCredit API key and connect your AI model API keys

    Sign up, open Developer → API keys, and generate a zc_… key. Then connect your AI model API keys in ZeroCredit — any mix of OpenAI, Anthropic, Google, xAI, DeepSeek, Kimi, Perplexity or Azure. You keep your existing billing relationships, discounts and rate limits; ZeroCredit never substitutes its own keys.

  2. 2

    Point your client, or send from ZeroCredit

    Choose the path that fits your stack. Point your client, framework or agent at ZeroCredit and keep every model name as-is, or make requests directly from ZeroCredit using your connected provider keys:

    https://zerocreditai.com/api/public/v1
  3. 3

    Send traffic — that's it

    Your first request appears in the dashboard within seconds. Semantic caching, budget controls, fallback chains, intelligent routing and ML cost optimization activate automatically; every claimed saving traces to a request ID.

# works with any standard chat client (openai, litellm, langchain, llama-index…)
client = ChatClient(
    api_key="zc_your_key",                       # Your ZeroCredit API key; providers stay on your own accounts
    base_url="https://zerocreditai.com/api/public/v1",  # ← the only line that changes
)

resp = client.chat.completions.create(
    model="gemini-2.5-pro",                      # keep any provider's model name, or "auto"
    messages=[{"role": "user", "content": "Hello"}],
)
print(resp.choices[0].message.content)

Compatibility checklist

What works when you route any provider's models through ZeroCredit.

Chat completions (non-streaming)
Streaming (SSE)
JSON mode / structured responses
Function / tool calling— routed only to providers that support tools
temperature, max tokens, top_p and friends
Any provider's model name — GPT, Claude, Gemini, Grok, DeepSeek, Kimi, Azure— routed to your own provider keys
Logical models: "auto", "cheapest", "fastest"— ZeroCredit picks the best model per request
Bring your own provider keys (BYOK)— your existing accounts, discounts and rate limits stay in place
Semantic caching— automatic cache hits for similar prompts
Budget controls— hard-stop or alert-only spend guards
Fallback chains— automatic provider/model failover
Per-request savings proof— cost and model headers on every response
Multimodal content— inline images and document analysis pass through
Embeddings endpoint— used internally for caching; not exposed yet
Provider-specific stateful APIs (assistants, threads, files)— keep those calls on the provider's own SDK
Fine-tuning APIs
Create your key

Not ready to switch? Run the free cost audit on your current bill first.