> ## Documentation Index
> Fetch the complete documentation index at: https://docs.inceptionlabs.ai/llms.txt
> Use this file to discover all available pages before exploring further.

# Quickstart

> Get started with Mercury 2 for voice applications.

*Check out our guides for creating voice agents with Mercury in [ElevenLabs](/cookbooks/elevenlabs-agent) and [Vapi](/resources/vapi).*

Mercury plugs in easily to ASR → LLM → TTS voice agent workflows, providing quality and latencies that are competitive with non-reasoning small frontier models and reasoning models served on specialized hardware.

Set your Inception API key:

<CodeGroup>
  ```bash macOS / Linux theme={null}
  export INCEPTION_API_KEY="your_api_key_here"
  ```

  ```bash Windows theme={null}
  set INCEPTION_API_KEY="your_api_key_here"
  ```
</CodeGroup>

Install the SDK:

<CodeGroup>
  ```bash Python theme={null}
  pip install inceptionai
  ```

  ```bash TypeScript theme={null}
  npm install inceptionai
  ```
</CodeGroup>

Try a request:

<CodeGroup>
  ```bash cURL theme={null}
  curl https://api.inceptionlabs.ai/v1/chat/completions \
    -H "Content-Type: application/json" \
    -H "Authorization: Bearer $INCEPTION_API_KEY" \
    -d '{
      "model": "mercury-2",
      "messages": [
        {"role": "system", "content": "You are speaking in a voice interface. Follow these rules strictly:\n1. NEVER use markdown, asterisks, bold, bullet points, numbered lists, or headers.\n2. Keep every response as short as possible.\n3. If a tool requires missing information, ask for ONE thing at a time."},
        {"role": "user", "content": "Hey, what'\''s up?"}
      ],
      "reasoning_effort": "medium",
      "temperature": 0.75,
      "max_tokens": 8192,
      "realtime": true
    }'
  ```

  ```python Python theme={null}
  from inceptionai import Inception

  client = Inception()  # reads INCEPTION_API_KEY from the environment

  SYSTEM_PROMPT = (
      "You are speaking in a voice interface. Follow these rules strictly:\n"
      "1. NEVER use markdown, asterisks, bold, bullet points, numbered lists, or headers.\n"
      "2. Keep every response as short as possible.\n"
      "3. If a tool requires missing information, ask for ONE thing at a time."
  )

  completion = client.chat.completions.create(
      model="mercury-2",
      messages=[
          {"role": "system", "content": SYSTEM_PROMPT},
          {"role": "user", "content": "Hey, what's up?"},
      ],
      reasoning_effort="medium",
      temperature=0.75,
      max_tokens=8192,
      realtime=True,
  )
  print(completion.choices[0].message.content)
  ```

  ```typescript TypeScript theme={null}
  import Inception from 'inceptionai';

  const client = new Inception(); // reads INCEPTION_API_KEY from the environment

  const SYSTEM_PROMPT = [
    'You are speaking in a voice interface. Follow these rules strictly:',
    '1. NEVER use markdown, asterisks, bold, bullet points, numbered lists, or headers.',
    '2. Keep every response as short as possible.',
    '3. If a tool requires missing information, ask for ONE thing at a time.',
  ].join('\n');

  const completion = await client.chat.completions.create({
    model: 'mercury-2',
    messages: [
      { role: 'system', content: SYSTEM_PROMPT },
      { role: 'user', content: "Hey, what's up?" },
    ],
    reasoning_effort: 'medium',
    temperature: 0.75,
    max_tokens: 8192,
    realtime: true,
  });
  console.log(completion.choices[0]?.message.content);
  ```
</CodeGroup>

We recommend `temperature=0.75` (default) and `reasoning_effort="medium"` for Mercury in most voice use cases. For tool-calling workflows, a lower temperature may be appropriate, e.g., `temperature=0.6`.

You can also try out `reasoning_effort="low"` and `realtime=true` for ultra-low latency, especially TTFT-sensitive workflows.

Since Mercury is a reasoning model, we recommend setting `max_tokens` to at least 3,000 for `reasoning_effort="medium"`, adjusting as needed according to reasoning effort.

| Priority                          | temperature | reasoning\_effort | realtime |
| --------------------------------- | ----------- | ----------------- | -------- |
| Standard voice agent              | 0.75        | medium            | true     |
| Time to First Answer Token (TTFT) | 0.75        | low               | true     |