10-Minute API Quickstart

Use ACP as an API in your applications. This guide walks you through authentication, your first request, response interpretation, and practical use cases.

Step 1: Get your API key

ACP uses OpenRouter to access multiple LLM providers through a single key. OpenRouter gives you access to models from OpenAI, Anthropic, Google, Meta, and others.

1. Sign up at openrouter.ai
2. Navigate to openrouter.ai/keys and create a new API key
3. Export the key in your terminal session

# OpenRouter provides access to multiple LLMs through one key
# Sign up at: https://openrouter.ai/keys
export OPENROUTER_API_KEY="your_key_here"

Step 2: Make your first request

The consensus endpoint accepts a query and a list of models. It runs iterative consensus rounds until the D-score drops below the threshold or the maximum number of iterations is reached.

curl -X POST https://your-worker.workers.dev/consensus-iterative \
  -H "Content-Type: application/json" \
  -H "x-openrouter-key: $OPENROUTER_API_KEY" \
  -d '{
    "query": "What is 2+2?",
    "models": ["openai/gpt-5.4", "anthropic/claude-sonnet-4-6"],
    "max_iterations": 7
  }'

Request parameters

Step 3: Understand the response

A successful consensus response contains the final answer, convergence metrics, and a full iteration history showing how each model responded at every round.

{
  "query": "What is 2+2?",
  "final_answer": "4",
  "final_D": 0.0,
  "consensus_reached": true,
  "iterations_used": 1,
  "iteration_history": [
    {
      "iteration": 1,
      "D": 0.0,
      "responses": [
        { "model": "gpt-5.4", "content": "2+2 equals 4" },
        { "model": "claude-sonnet-4-6", "content": "The sum of 2 and 2 is 4" }
      ]
    }
  ]
}

Response fields

Example use cases

Below are ready-to-use examples for common integration patterns. Copy any of these and replace the API key and worker URL with your own.

Fact-checking

Use three or more models to verify factual claims. The consensus result is far more reliable than any single model's answer because hallucinations from one model are corrected by the others.

curl -X POST https://your-worker.workers.dev/consensus-iterative \
  -H "Content-Type: application/json" \
  -H "x-openrouter-key: $OPENROUTER_API_KEY" \
  -d '{
    "query": "What is the speed of light in vacuum?",
    "models": [
      "openai/gpt-5.4",
      "anthropic/claude-sonnet-4-6",
      "google/gemini-3.1-pro"
    ]
  }'

Code review

Submit code snippets for multi-model review. Each model independently analyzes the code, and ACP synthesizes their findings into a unified assessment.

curl -X POST https://your-worker.workers.dev/consensus-iterative \
  -H "Content-Type: application/json" \
  -H "x-openrouter-key: $OPENROUTER_API_KEY" \
  -d '{
    "query": "Is this Python code correct?\ndef factorial(n):\n    if n == 0: return 1\n    return n * factorial(n-1)",
    "models": ["openai/gpt-5.4", "anthropic/claude-sonnet-4-6"]
  }'

Content moderation

Use consensus to make more balanced moderation decisions. Multiple models evaluate content independently, reducing the bias that can come from relying on a single model's judgment.

curl -X POST https://your-worker.workers.dev/consensus-iterative \
  -H "Content-Type: application/json" \
  -H "x-openrouter-key: $OPENROUTER_API_KEY" \
  -d '{
    "query": "Evaluate this user comment for policy violations: [comment text here]",
    "models": [
      "openai/gpt-5.4",
      "anthropic/claude-sonnet-4-6",
      "google/gemini-3.1-pro"
    ],
    "max_iterations": 3
  }'

Choosing models

The models you choose affect cost, speed, and consensus quality. Here are recommended combinations for different scenarios:

Next steps

• Full Setup (30 min) -- run ACP locally with the complete frontend, Python engine, and Worker API.
• Worker API Reference -- full documentation of all endpoints, parameters, and response formats.
• Authentication -- detailed guide to API keys, rate limits, and security best practices.
• Code Examples -- Python, JavaScript, and cURL examples for common integration patterns.