ainative-openai

Drop-in replacement for the openai npm package — routes to AINative's free OpenAI-compatible API.

Get free access to Llama 3.3 70B, Qwen, DeepSeek, and more open-source models. Same API, zero config.

Quick Start

npm install ainative-openai openai

import OpenAI from 'ainative-openai';

const client = new OpenAI();

const response = await client.chat.completions.create({
  model: 'llama-3.3-70b',
  messages: [{ role: 'user', content: 'Explain quantum computing in one paragraph.' }]
});

console.log(response.choices[0].message.content);

That's it. No API key needed on first run — one is auto-provisioned for you.

Why?

| Feature | openai | ainative-openai | |---------|----------|-------------------| | Price | Pay per token | Free tier included | | Models | OpenAI only | Llama 3.3 70B, Qwen, DeepSeek, + more | | Setup | API key required | Zero config (auto-provisions) | | API | OpenAI API | Same OpenAI API | | Streaming | Yes | Yes | | Tool calling | Yes | Yes | | TypeScript | Yes | Yes (inherits from openai) |

How It Works

ainative-openai wraps the official openai npm package and:

1. Points baseURL to AINative's OpenAI-compatible API (https://api.ainative.studio/api/v1)
2. Auto-provisions a free account on first use if no API key is found
3. Saves credentials to ~/.ainative/config.json, .env, and .mcp.json

Everything else — streaming, tool calling, TypeScript types, error handling — comes directly from the openai package.

Available Models (Free Tier)

| Model | Best For | |-------|----------| | llama-3.3-70b | General purpose, reasoning | | qwen3-coder | Code generation | | deepseek-coder-v2 | Code + reasoning | | qwen-2.5-72b | Multilingual, math | | llama-3.1-8b | Fast, lightweight tasks |

Streaming

import OpenAI from 'ainative-openai';

const client = new OpenAI();

const stream = await client.chat.completions.create({
  model: 'llama-3.3-70b',
  messages: [{ role: 'user', content: 'Write a haiku about code.' }],
  stream: true,
});

for await (const chunk of stream) {
  process.stdout.write(chunk.choices[0]?.delta?.content || '');
}

Tool Calling

import OpenAI from 'ainative-openai';

const client = new OpenAI();

const response = await client.chat.completions.create({
  model: 'qwen3-coder',
  messages: [{ role: 'user', content: 'What is the weather in San Francisco?' }],
  tools: [{
    type: 'function',
    function: {
      name: 'get_weather',
      description: 'Get current weather for a city',
      parameters: {
        type: 'object',
        properties: {
          city: { type: 'string', description: 'City name' }
        },
        required: ['city']
      }
    }
  }]
});

console.log(response.choices[0].message.tool_calls);

Authentication

Credentials are resolved in this order:

1. opts.apiKey passed to constructor
2. AINATIVE_API_KEY environment variable
3. OPENAI_API_KEY environment variable
4. ZERODB_API_KEY environment variable
5. ~/.ainative/config.json
6. .mcp.json in project directory (walks up 6 levels)
7. Auto-provision a free account (credentials saved for next time)

Using an existing API key

export AINATIVE_API_KEY=ak_your_key_here

Or pass it directly:

const client = new OpenAI({ apiKey: 'ak_your_key_here' });

Claiming your auto-provisioned account

On first run, you'll see a claim URL printed to stderr:

ainative-openai: No API key found — provisioning a free account...
Auto-provisioned! Project: proj_abc123
Claim your account: https://ainative.studio/claim/abc123

Visit the claim URL to set a password and get persistent access.

Migrating from openai

- import OpenAI from 'openai';
+ import OpenAI from 'ainative-openai';

  const client = new OpenAI();
- // Requires OPENAI_API_KEY and costs money
+ // Free Llama/Qwen/DeepSeek — auto-provisions if no key set

  const res = await client.chat.completions.create({
-   model: 'gpt-4',
+   model: 'llama-3.3-70b',
    messages: [{ role: 'user', content: 'Hello!' }]
  });

Async Provisioning

If you need to ensure the API key is ready before making requests:

const client = new OpenAI();
await client.waitForProvisioning();
// Now safe to make requests

MCP Integration

After auto-provisioning, a .mcp.json entry is written so MCP-compatible tools (Claude Code, Cursor, Windsurf) can use your credentials:

{
  "mcpServers": {
    "ainative-openai": {
      "command": "npx",
      "args": ["-y", "ainative-openai"],
      "env": {
        "AINATIVE_API_KEY": "ak_...",
        "AINATIVE_PROJECT_ID": "proj_..."
      }
    }
  }
}

License

MIT

Links

• AINative Studio
• API Documentation
• GitHub
• npm