SandBase

One API. 2,000+ Models. 2,000+ APIs. Zero integration headaches.

If SandBase saves you integration time, please star the repository — it helps more builders discover the project.

The Problem

Building AI applications means juggling dozens of API keys, learning different SDKs, handling auth flows, managing rate limits, and paying for 10+ subscriptions just to get basic capabilities.

Need Twitter data? One API. Web scraping? Another. Image generation? A third. LLM inference? Yet another. Each with its own pricing, auth, error handling, and documentation.

What if one API key could access all of them?

The Solution

SandBase is a unified gateway to 2,000+ AI models and 2,000+ APIs through a single interface:

# Call any LLM — GPT-4o, Claude, Gemini, DeepSeek, Qwen...
curl https://sandbase.ai/v1/run -d '{"model": "openai/gpt-4o", "messages": [...]}'

# Search Twitter
curl https://sandbase.ai/v1/run -d '{"model": "twitter/search", "query": "AI agents"}'

# Generate an image with Flux
curl https://sandbase.ai/v1/run -d '{"model": "flux/schnell", "prompt": "A startup logo"}'

# Scrape any webpage
curl https://sandbase.ai/v1/run -d '{"model": "firecrawl/scrape", "url": "https://..."}'

# Generate a video with Kling
curl https://sandbase.ai/v1/run -d '{"model": "kling/video", "prompt": "Product demo"}'

One endpoint. One API key. One billing dashboard. That's it.

2,000+ Models

Type	Count	Providers
LLM	500+	OpenAI (GPT-4o, o1, o3), Anthropic (Claude 4, Sonnet), Google (Gemini 2.5), DeepSeek (V3, R1), Qwen (3), Meta (Llama 4), Mistral, Cohere, Yi...
Image Generation	200+	Flux (Schnell, Pro, Dev), DALL-E 3, Ideogram 3, Recraft V3, Stable Diffusion 3.5, Midjourney...
Video Generation	100+	Kling 2.0, MiniMax, Runway Gen-4, Luma Dream Machine, Pika, Hailuo...
Audio	100+	ElevenLabs (TTS/clone), OpenAI Whisper, Fish Audio, Suno, Udio...
Embeddings	50+	OpenAI (text-3-large), Voyage 3, Cohere Embed V4, Jina V3...
Multimodal	200+	GPT-4o vision, Gemini Pro Vision, Claude vision, Qwen-VL...

2,000+ APIs

Category	Count	Examples
Web Search	50+	Google, Exa, Tavily, Serper, Scholar, Brave, DuckDuckGo
Social Media	300+	Twitter/X (search, timeline, user), YouTube (transcript, search), Reddit, Instagram, TikTok, LinkedIn, Xiaohongshu, Weibo, Bilibili, Douyin
Web Scraping	100+	Firecrawl (crawl, scrape, map), Exa contents, Jina Reader, URL extract
E-commerce	200+	Taobao (search, detail), Amazon, AliExpress, JD, Pinduoduo
Finance	100+	Stock quotes, company fundamentals, SEC filings, forex rates
Maps & Weather	50+	Google Maps, weather forecasts, geocoding, places
Translation	50+	DeepL, Google Translate, specialized domain translation
News	50+	Google News, Bing News, RSS aggregation, media monitoring
Developer Tools	100+	GitHub, code search, package info, DNS/WHOIS, SSL check
And more	500+	Travel, food, sports, education, government data, academic papers...

See It In Action

Your AI agent searches the web in real-time

You: "What happened with OpenAI this week?"

Agent → sandbase_discover("web search") → sandbase_run("exa/search")
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
→ 10 results from the past 7 days:
  1. "OpenAI Announces GPT-5 Preview" — The Verge (2 days ago)
  2. "New API Pricing for o3-mini" — OpenAI Blog (3 days ago)
  ...

Generate images on demand

You: "Create a logo for my app called 'Moonshot' — dark theme, minimal"

Agent → sandbase_run("flux/schnell", {prompt: "..."})
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
✓ Generated: moonshot-logo.png (1024x1024)
  Cost: $0.003 | Time: 1.8s

You: "Find what people are saying about our competitor on Twitter"

Agent → sandbase_run("twitter/search", {query: "CompetitorName"})
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
→ 25 recent posts mentioning CompetitorName:
  Sentiment: 60% positive, 25% neutral, 15% negative
  Top complaint: "pricing increased 40%"
  Top praise: "new AI features are solid"

Scrape and analyze any website

You: "Get the pricing from linear.app and compare with our plans"

Agent → sandbase_run("firecrawl/scrape", {url: "https://linear.app/pricing"})
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
✓ Extracted 3 plans:
  Free: $0/mo — 250 issues
  Standard: $8/user/mo — unlimited
  Plus: $14/user/mo — advanced analytics

Three Ways to Use SandBase

1. Direct API — Full control

python

import requests

# Call any model with one unified API
response = requests.post("https://sandbase.ai/v1/run",
    headers={"Authorization": "Bearer YOUR_KEY"},
    json={"model": "openai/gpt-4o", "messages": [...]}
)

# Or use OpenAI-compatible endpoint (drop-in replacement)
response = requests.post("https://sandbase.ai/v1/chat/completions",
    headers={"Authorization": "Bearer YOUR_KEY"},
    json={"model": "openai/gpt-4o", "messages": [...]}
)

2. CLI — Connect any AI coding agent in 30 seconds

npx -y @sandbaseai/cli connect

Your Cursor / Claude Code / Codex / Kiro / Windsurf instantly gets access to all 2,000+ models and 2,000+ APIs through MCP. Just ask naturally.

Supports 17+ clients: Cursor, Claude Code, Codex, Kiro IDE, Kiro CLI, Windsurf, Gemini CLI, Amp, Warp, OpenCode, Qwen Code, Kimi CLI, Hermes, OpenClaw, ChatGPT, Claude Desktop, and more.

3. Managed Agents — Reusable AI workflows

json

{
  "name": "daily-market-report",
  "model": "openai/gpt-4o",
  "instructions": "Research market trends, analyze competitor moves, summarize findings",
  "tools": ["twitter/search", "exa/search", "firecrawl/scrape"],
  "schedule": "0 9 * * 1-5"
}

Define once, run on schedule, invoke via API, or trigger by webhook.

Getting Started (under 2 minutes)

Sign up at sandbase.ai — free tier included
Get your API key from the Dashboard
Make your first call:

curl https://sandbase.ai/v1/chat/completions \
  -H "Authorization: Bearer $SANDBASE_KEY" \
  -H "Content-Type: application/json" \
  -d '{"model": "openai/gpt-4o", "messages": [{"role": "user", "content": "Hello!"}]}'

Or connect your AI tool:

npx -y @sandbaseai/cli connect --client cursor

Why SandBase?

Without SandBase	With SandBase
20+ API keys to manage	1 API key for everything
Different SDKs per service	One unified REST API
$500+/mo in separate subscriptions	Pay only for what you use
Weeks of integration per service	Working in 2 minutes
Each provider has different errors	Consistent error format
Separate billing per provider	One dashboard, one invoice
Manual rate limit handling	Built-in retry and fallback
No visibility across providers	Unified logs, costs, and analytics

Documentation

This repository contains the full SandBase documentation site, built with VitePress.

Section	What's Inside
Getting Started	Account setup, API keys, first call
Models	2,000+ models — capabilities, pricing
API Reference	Complete REST API docs
Agents	Build and deploy managed agents
Guides	Streaming, errors, rate limits, billing
For AI Tools	MCP integration for coding agents
Store	Browse all available models & APIs

Run locally

npm ci
npm run dev      # Dev server at localhost:5173
npm run build    # Production build
npm run preview  # Preview build locally

Deploy

docker build -t sandbase-docs .
docker run -p 80:80 sandbase-docs

License

Apache-2.0

Contributing

Found a missing model, unclear example, or documentation bug? Open an issue or send a pull request. Small fixes are welcome.

SandBase

The Problem ​

The Solution ​

2,000+ Models ​

2,000+ APIs ​

See It In Action ​

Your AI agent searches the web in real-time ​

Generate images on demand ​

Pull live social media data ​

Scrape and analyze any website ​

Three Ways to Use SandBase ​

1. Direct API — Full control ​

2. CLI — Connect any AI coding agent in 30 seconds ​

3. Managed Agents — Reusable AI workflows ​

Getting Started (under 2 minutes) ​

Why SandBase? ​

Documentation ​

Run locally ​

Deploy ​

Links ​

License ​

Contributing ​