Documentation
Guides / Quickstart
Guides · Quickstart

Quickstart

Get started with AllToken.

Introduction

AllToken provides a unified API with access to leading AI models through a single endpoint, with automatic fallbacks and cost-effective routing built in.

Get started in minutes with your preferred SDK or HTTP client.

Base URL: https://api.alltoken.ai/v1

Auth: Bearer API key

Compatibility: OpenAI-compatible API

Get your API key

Before getting started, create an API key:

  1. Go to Settings → API Keys
  2. Click Create new key
  3. Copy and save the key securely — it's shown only once

Keep your API key secret. Do not expose it in client-side code or public repositories.

Install as an agent skill

Skip writing integration code if your stack already speaks AllToken. AllToken ships two official skills for agent runtimes that load SKILL.md files - install them in one command and your agent learns the entire AllToken API surface.

SkillWhat it does
alltokenBootstrap a complete TypeScript or Python AllToken project: chat, async image, async video, streaming, tool-calling agent core, and optional Ink TUI.
alltoken-callSix slash-style commands the agent recognizes in chat: /alltoken-chat, /alltoken-image, /alltoken-video, /alltoken-search, /alltoken-models, /alltoken-cost. Stdlib Python recipes - no pip install.

Both ship under MIT. Source: github.com/alltoken-ai/alltoken-skills.

Install in your runtime

OpenClaw (via the ClawHub CLI):

OpenClaw
$clawhub skill install alltoken
clawhub skill install alltoken-call

Hermes Agent (direct URL install):

Hermes Agent
$hermes skills install https://alltoken.ai/skills/alltoken/SKILL.md
hermes skills install https://alltoken.ai/skills/alltoken-call/SKILL.md

Claude Code / Codex CLI / OpenCode - drop the SKILL.md file into your project's skills/ directory:

Claude Code / Codex CLI / OpenCode
$mkdir -p ./skills/alltoken-call
curl -fsSL https://alltoken.ai/skills/alltoken-call/SKILL.md -o ./skills/alltoken-call/SKILL.md

Your agent will pick up the skill on the next session.

Try it

Make sure your ALLTOKEN_API_KEY is set in the environment your agent runs in. Then ask in natural language:

  • "Use the alltoken-call skill to generate a 1024x1024 image of a teapot."
  • "Use alltoken to scaffold an AllToken chat project in ./my-agent."
  • "Run /alltoken-models --type=video to show available video models."

Using SillyTavern for roleplay? See the dedicated guide: Use AllToken with SillyTavern.

Prefer to write your own integration? Continue to Install the SDK below for the TypeScript / Python paths.

Install the SDK

Use the OpenAI SDK with AllToken. Install it with your preferred package manager:

npm
$npm install openai

Then set your environment variable:

Shell
$export ALLTOKEN_API_KEY="your_alltoken_api_key"

Send your first request

Create a client, pick a model, and send a chat completion:

TypeScript
1import OpenAI from 'openai';
2
3const client = new OpenAI({
4 apiKey: process.env.ALLTOKEN_API_KEY,
5 baseURL: 'https://api.alltoken.ai/v1',
6});
7
8const completion = await client.chat.completions.create({
9 model: 'minimax-m2.7',
10 messages: [
11 {
12 role: 'user',
13 content: 'What is the meaning of life?',
14 },
15 ],
16});
17
18console.log(completion.choices[0]?.message?.content);

Python example

Python
1from openai import OpenAI
2import os
3
4client = OpenAI(
5 api_key=os.environ.get("ALLTOKEN_API_KEY"),
6 base_url="https://api.alltoken.ai/v1",
7)
8
9completion = client.chat.completions.create(
10 model="minimax-m2.7",
11 messages=[
12 {"role": "user", "content": "What is the meaning of life?"}
13 ],
14)
15
16print(completion.choices[0].message.content)

Using the API directly

Call the API directly with cURL or any HTTP client:

cURL
1curl https://api.alltoken.ai/v1/chat/completions \
2 -H "Authorization: Bearer $ALLTOKEN_API_KEY" \
3 -H "Content-Type: application/json" \
4 -d '{
5 "model": "minimax-m2.7",
6 "messages": [
7 {"role": "user", "content": "Hello!"}
8 ]
9 }'

Streaming responses

Add stream: true to get responses token-by-token via Server-Sent Events:

TypeScript
1const stream = await client.chat.completions.create({
2 model: 'minimax-m2.7',
3 messages: [{ role: 'user', content: 'Tell me a story' }],
4 stream: true,
5});
6
7for await (const chunk of stream) {
8 const content = chunk.choices[0]?.delta?.content;
9 if (content) process.stdout.write(content);
10}

For detailed streaming documentation, see Streaming.

Next steps