Access 170+ open-source AI models, intelligent search endpoints, and pre-built agentic workflows through a single, OpenAI-compatible API. Built on Cloudflare's global network for reliability at scale.
Built-in OAuth, sub-tenant management, and payment integration. Zero config, just code.
# NPM
npm install @frnd/uplink-sdk
# or curl (no dependencies)
curl -fsSL https://api.frnds.cloud/sdk/install.sh | bashimport { initializeDeveloperAccount }
from '@frnd/uplink-sdk'
await initializeDeveloperAccount({
appName: 'My App',
plan: 'pro'
})
// Opens browser, creates sub-tenant, saves credentialsimport { getUplinkClient }
from '@frnd/uplink-sdk'
const uplink = await getUplinkClient()
const response = await uplink.chat({
model: 'llama-3.3-70b-versatile',
messages: [{ role: 'user', content: 'Hello!' }]
})Browser-based auth with device code flow. No password storage.
Isolated tenancy per app. Automatic credential management.
Stripe/Paddle integration. Validate payment before activation.
Auto-loads credentials. No environment variables needed.
Pre-configured endpoints and agentic workflows that combine models with real-world data and tools.
Never break production again. Automatic model migration with full transparency when providers deprecate models.
// Auto-resolves deprecated models
X-Model-Deprecated: true
Stop memorizing exact model names. 80+ aliases handle dashes, underscores, prefixes, and typos automatically.
// All work: llama-3.3-70b, llama_3_3_70b
Latest multimodal models with vision, tool calling, and 128K context. Scout (17B, 16 experts) and Maverick (17B, 128 experts).
meta-llama/llama-4-scout-17b-16e-instruct
Real-time web search with AI synthesis. Get answers, not just links.
magic.search('latest quantum breakthroughs')
Extract and understand content from any URL with intelligent parsing.
magic.extract('https://arxiv.org/...')
Pre-built multi-step workflows for complex tasks like research and analysis.
magic.workflow('market-research', params)
Extract structured data and insights from unstructured content.
magic.analyze('financial-report.pdf')
Upload documents, create embeddings, and build semantic search with human-readable citations. Documents cited as "Employee-Handbook.pdf" instead of hex IDs.
POST /evs/sources { "documents": [...] }
Intelligent document compression with 10-100x reduction. Auto-routing picks the optimal method for your content.
POST /auto/ingest (EVS|OCR|OCV)
Production TTS with telephony support. Drop-in replacement for ElevenLabs with µ-law encoding for phone systems.
POST /v1/text-to-speech/{voice_id}
Production-ready text-to-speech with telephony support. Drop-in replacement for ElevenLabs with 30% lower costs and µ-law encoding for phone systems.
# Convert text to speech
curl -X POST https://api.frnds.cloud/v1/text-to-speech/uplink_rachel_001 \
-H "Authorization: Bearer YOUR_KEY" \
-H "Content-Type: application/json" \
-d '{
"text": "Hello from Uplink Voice API",
"output_format": "mp3_44100_128"
}' --output voice.mp3
5 premium voices, 20+ languages, streaming support
# Generate telephony-grade audio
curl -X POST https://api.frnds.cloud/v1/text-to-speech/uplink_rachel_001 \
-H "Authorization: Bearer YOUR_KEY" \
-H "Content-Type: application/json" \
-d '{
"text": "Press 1 for sales, 2 for support",
"output_format": "mp3_8000_16" // 8kHz for phone
}' --output phone.mp3
MP3 8kHz/16kHz formats, < 150ms latency
# AI-powered voice responses
curl -X POST https://api.frnds.cloud/v1/voice/agent \
-H "Authorization: Bearer YOUR_KEY" \
-H "Content-Type: application/json" \
-d '{
"messages": [{"role": "user", "content": "Hello"}],
"voice_config": {
"voice_id": "uplink_rachel_001",
"emotion": "friendly"
},
"enable_voice": true
}'
Combine LLM intelligence with natural speech
Complete access to Llama, Qwen, Mistral, and 170+ open-source models. No vendor lock-in, just pure model access.
Create sub-tenants for your users with isolated data, custom quotas, and independent billing. Perfect for SaaS platforms.
Intelligent routing automatically selects the most cost-effective provider while maintaining quality.
Pre-built multi-step workflows for research, analysis, and complex reasoning tasks. No orchestration needed.
Deployed across Cloudflare's 300+ locations. Sub-50ms latency for 95% of internet users.
Drop-in replacement works with existing OpenAI SDKs and libraries. Zero code changes required.
Production-ready text-to-speech with telephony support. Drop-in replacement for ElevenLabs with µ-law encoding for phone systems.
Build RAG applications with semantic search. Upload documents, create embeddings, and query with natural language.
Start free, upgrade when you need more. All plans include access to 170+ models.
Perfect for testing
For side projects
Most popular
For production apps
Access the latest open-source AI models from Llama, Qwen, Mistral, and more through one unified API.
Built on Cloudflare's global network, Uplink delivers consistent performance whether you're serving 10 or 10 million users.