Route AI calls through one metered proxy.

1. Register or log in at MeterLayer.
2. Add your OpenAI key in Settings - Providers.
3. Create a project in Projects.
4. Copy the generated project API key.
5. Use baseURL: https://meterlayer.io/v1 in OpenAI-compatible clients.
6. Run a low-token test in Dashboard Playground.

Machine-readable API spec: OpenAPI JSON.

Proxy calls use a project API key, not a user session. Keep this key on your server. Do not expose it in browser JavaScript, mobile apps, or public repos.

Authorization: Bearer <project_api_key>
Content-Type: application/json

Existing OpenAI clients can use the gateway by changing the base URL and API key. The project API key goes into the normal OpenAI SDK apiKey field.

POST https://meterlayer.io/v1/chat/completions

import OpenAI from "openai";

const client = new OpenAI({
  apiKey: process.env.AI_COST_GATEWAY_PROJECT_KEY,
  baseURL: "https://meterlayer.io/v1"
});

const completion = await client.chat.completions.create({
  model: "gpt-4o-mini",
  messages: [{ role: "user", content: "Reply with one word: OK" }],
  max_tokens: 5,
  temperature: 0.2,
  metadata: {
    feature: "first-smoke-test",
    customerId: "internal-test"
  }
});

The endpoint accepts model,messages,max_tokens, generation options such as temperature and optionalmetadata. Streaming is not supported yet.

The gateway validates the project key, checks project status, model allow-list, budget, rate limit and provider credential, then forwards the request to OpenAI.

{
  "provider": "openai",
  "model": "gpt-4o-mini",
  "messages": [
    { "role": "user", "content": "Hello" }
  ],
  "maxOutputTokens": 1024,
  "temperature": 0.2,
  "topP": 0.9,
  "metadata": {
    "userId": "optional-user-id",
    "feature": "email-generator",
    "customerId": "optional-customer-id"
  }
}

On success, the body is the provider response. MeterLayer adds response headers with request ID, estimated cost and rate-limit information.

Rate-limit headers are included after the project API key is validated and the project rate-limit check has run.

{
  "id": "chatcmpl_...",
  "object": "chat.completion",
  "model": "gpt-4o-mini",
  "choices": [
    {
      "message": {
        "role": "assistant",
        "content": "Hello! How can I help?"
      },
      "finish_reason": "stop"
    }
  ],
  "usage": {
    "prompt_tokens": 9,
    "completion_tokens": 7,
    "total_tokens": 16
  }
}

curl

curl -X POST https://meterlayer.io/api/proxy/chat \
  -H "Authorization: Bearer $AI_COST_GATEWAY_PROJECT_KEY" \
  -H "Content-Type: application/json" \
  -H "X-Request-Id: demo-request-001" \
  -d '{
    "provider": "openai",
    "model": "gpt-4o-mini",
    "messages": [
      { "role": "system", "content": "You write concise product copy." },
      { "role": "user", "content": "Write one sentence about an AI cost dashboard." }
    ],
    "maxOutputTokens": 200,
    "metadata": {
      "feature": "product-description",
      "customerId": "customer-123"
    }
  }'

TypeScript

const response = await fetch("https://meterlayer.io/api/proxy/chat", {
  method: "POST",
  headers: {
    Authorization: `Bearer ${process.env.AI_COST_GATEWAY_PROJECT_KEY}`,
    "Content-Type": "application/json",
    "X-Request-Id": crypto.randomUUID()
  },
  body: JSON.stringify({
    provider: "openai",
    model: "gpt-4o-mini",
    messages: [{ role: "user", content: "Hello" }],
    metadata: {
      userId: "user-123",
      feature: "chat"
    }
  })
});

const data = await response.json();
if (!response.ok) throw new Error(data.error);
return data;

Use this sequence for the first production feature routed through MeterLayer. Keep the project API key in a backend secret manager, set a conservative project budget, and restrict allowed models before sending traffic.

1. Add the customer OpenAI key in Settings - Providers and run the provider test.
2. Create a project for the app or environment, for example web-prod.
3. Copy the project API key once and store it server-side.
4. Send one low-token smoke request with metadata.feature.
5. Open Logs, Prompts, Models and Budgets.
6. Record X-AI-Cost-Gateway-Request-Id for support.

const completion = await client.chat.completions.create({
  model: "gpt-4o-mini",
  messages: [{ role: "user", content: "Reply with one word: OK" }],
  max_tokens: 16,
  metadata: {
    feature: "first-customer-smoke",
    customerId: "internal-test"
  }
});

Rollback is simple: restore the original OpenAI API key and base URL in the customer application. Support should ask for timestamp, project, model, feature and request ID, not provider keys, project keys, full prompts or personal data.

Send a unique X-Request-Id for retries and traceability. If the same project sends the same ID again, the gateway returns 409 Duplicate request ID.

Error responses include an error message andrequestId. Validation failures also include details.

{
  "error": "Provider or model is not allowed for this project.",
  "requestId": "req_..."
}

inputCost = inputTokens / 1_000_000 * inputPer1M
outputCost = outputTokens / 1_000_000 * outputPer1M
totalCost = inputCost + outputCost