Every model,one key.
Run leading models in Claude Code, Codex and Cursor.
Sepolia testnet
- Start with $2.50
How it works
Turn INFERENCE into credit.
- Why Deference
- An OpenAI-compatible API routed through OpenRouter. Buy INFERENCE with USDC, then activate it into credit.
- What requests cost
- OpenRouter's price plus OpenRouter's 5% platform fee. Deference adds no markup.
- Why hold the launch token
- Trading fees fund INFERENCE for eligible launch token holders. Claim tokens or credit.
USDC in · INFERENCE out
USDC buys INFERENCE
The selected purchase route delivers INFERENCE. Its quote includes the purchase fees.
Burn tokens into credit
Activation burns INFERENCE. The amount after the activation fee becomes service credit that cannot be sent, sold or withdrawn.
Hold, then charge
A request holds its prompt plus max_tokens. When it ends, the hold is released and the measured cost is charged.
Connect
Connect your AI tools.
Set your API key and base URL in Claude Code, Codex or Cursor.
~/.zshrc
export ANTHROPIC_BASE_URL="https://deference.dea.web.id"export ANTHROPIC_AUTH_TOKEN="$DEFERENCE_API_KEY"export CLAUDE_CODE_ENABLE_GATEWAY_MODEL_DISCOVERY="1"
~/.codex/config.toml
model = "openai/gpt-5.3-codex"model_provider = "deference"web_search = "disabled"[model_providers.deference]name = "Deference"base_url = "https://deference.dea.web.id/v1"wire_api = "responses"env_key = "DEFERENCE_API_KEY"
Settings › Models › API Keys
opencode.json
{"provider": {"deference": {"npm": "@ai-sdk/openai-compatible","name": "Deference","options": {"baseURL": "https://deference.dea.web.id/v1","apiKey": "{env:DEFERENCE_API_KEY}"}}},"model": "deference/anthropic/claude-sonnet-5.5"}
app.py
import osfrom openai import OpenAIclient = OpenAI(base_url="https://deference.dea.web.id/v1",api_key=os.environ["DEFERENCE_API_KEY"],)
One API
- Chat Completions, Responses and Messages
- Streaming
- Tool calls and structured outputs
- Docs your agent can read
Control
Set limits. Track requests.
Give each tool its own key and credit limit. Inspect the model, token usage and final cost of every request.
Example: the production key has a $50.00 credit limit and has spent $41.20. One Claude Sonnet 5.5 request used 1,200 input tokens and 350 output tokens, with a final cost of $0.0062 charged to this key.
Example data
Launch token trades produce WETH fees. The configured split separates the holders’ share from the treasury.
Questions
Read the docsCan I use it today?
Yes, on the Sepolia testnet. Create a key, use up to $2.50 of free credit on open-weight models, and add credit with test USDC.
How is a request priced?
OpenRouter's price plus OpenRouter's 5% platform fee. Deference adds no markup.
How is this different from OpenRouter?
Pay for model requests with INFERENCE. Purchase and activation records are on chain, and DEFERENCE holders can claim rewards from trading fees.
Why hold DEFERENCE?
Each hour, available trading fees fund INFERENCE for eligible snapshot holders. Claim tokens or credit.
Can I get my money back out?
Credit cannot be withdrawn or sent. INFERENCE you have not activated can be sold on a DEX at the market price.
Do you store my prompts?
No. Deference keeps request metadata, such as model, tokens, latency and cost, not the content. OpenRouter and the model providers have their own policies.
Which tools work with Deference?
Claude Code, Codex, Cursor (beta), OpenCode, Cline, Zed, Continue, Goose and more, each with a setup guide. Gemini CLI does not; its request format is not supported.
Start with up to $2.50 of free credit.
Verify an email, create a key and run an open-weight model.
Example: the get-started checklist after the first three steps, with $1.00 of free credit added.