Use aiplug
Quick start

Quick start

Install

npm install aiplug

Configure a provider

The CLI is the fastest path:

# Interactive — walks you through apiKey, baseURL, model selection
npx aiplug transport add openai

# Or non-interactive
npx aiplug transport add anthropic --api-key=$ANTHROPIC_API_KEY --model=claude-3-5-sonnet-latest --force --yes
npx aiplug transport add ollama    --base-url=http://localhost:11434 --model=llama3.2 --force --yes

Aiplug stores config in ~/.config/aiplug/. The active provider is one CLI command away:

npx aiplug transport use anthropic
npx aiplug status

Run the HTTP server

npx aiplug serve --port=3711

This exposes an OpenAI-compatible API at http://127.0.0.1:3711. Point any OpenAI SDK at it:

import OpenAI from 'openai';

const client = new OpenAI({
  baseURL: 'http://localhost:3711/v1',
  apiKey: 'whatever',  // aiplug handles the real auth
});

const reply = await client.chat.completions.create({
  model: 'claude-3-5-sonnet-latest',
  messages: [{ role: 'user', content: 'Hello' }],
});

The application code never knows which provider is active. Switch providers with aiplug transport use <slug> and the same SDK calls route to a different backend.

Try the chat REPL

npx aiplug chat claude-3-5-sonnet-latest

Minimal streaming REPL. No banner, no onboarding, no colour noise. Direct prompt → streamed reply → next prompt. Slash commands: /help, /model <name>, /provider, /clear, /exit. Ctrl+C aborts the current stream; idle Ctrl+C or Ctrl+D exits.

Use it programmatically

import { AIPlug } from 'aiplug';

const ai = new AIPlug({
  transport: 'anthropic',
  apiKey: process.env.ANTHROPIC_API_KEY,
  model: 'claude-3-5-sonnet-latest',
});

const reply = await ai.chat({
  model: 'claude-3-5-sonnet-latest',
  messages: [{ role: 'user', content: 'Hello!' }],
});

console.log(reply.message.content);

For streaming, iterate the async iterable:

for await (const chunk of ai.stream({
  model: 'claude-3-5-sonnet-latest',
  messages: [{ role: 'user', content: 'Tell me a story.' }],
})) {
  if (chunk.type === 'text-delta') process.stdout.write(chunk.delta);
  if (chunk.type === 'finish') console.log('\n[done]', chunk.reason);
}

What just happened

Aiplug looked up the registry entry for anthropic in data/registry.json.
It dynamic-imported the transport module at dist/providers/anthropic/index.js.
The transport’s chat() made one fetch() call to the Anthropic API.
Aiplug normalised the response into the canonical ChatMessage / ChatResponse shape.

No hidden retries. No routing layer. No token accounting unless you ask for it.

Next steps

Embedding guide — drop aiplug into another project.
Provider registry — how providers get registered.
CLI reference — every aiplug command.