Built-in edge AI for platform features
SAM runs on Cloudflare Workers, and Workers AI provides edge inference for platform features. SAM uses Workers AI for task title generation, text-to-speech, and context summarization without requiring a separate model API key.
Why use Workers AI with SAM
No Separate Model Key
Workers AI is part of SAM's Cloudflare deployment — no additional model API key is needed for task titles and TTS.
Edge Inference
Models run at Cloudflare's edge — low latency for platform features, no cold starts.
Multiple Models
Llama 4, Gemma 4, GLM-4.7, Qwen 3, and Deepgram for different tasks — SAM picks the right model automatically.
AI Gateway
All inference is routed through Cloudflare AI Gateway for rate limiting, analytics, and reliability.
Get started in four steps
Already Built In
Workers AI is part of SAM's Cloudflare infrastructure — no setup required.
Automatic Model Selection
SAM uses the right model for each task: GLM for titles, Deepgram for TTS, Gemma for summarization.
AI Proxy (Optional)
Enable the AI proxy to route custom inference through Workers AI with rate limiting and token budgets.
Monitor Usage
Track AI usage, costs, and model distribution via the admin analytics dashboard.
What you can build
Zero-config AI features
Task title generation, text-to-speech for messages, and context summarization work out of the box.
AI proxy gateway
Route LLM requests through SAM's AI proxy for rate limiting, token budgets, and centralized analytics.
Self-hosted AI platform
When self-hosting SAM on the Workers Paid plan, Workers AI powers platform features without a separate model API key.
Frequently asked questions
Does Workers AI require a separate API key?
No — Workers AI is available through the Cloudflare account that runs SAM. Self-hosting SAM still requires the Workers Paid plan because the default deployment uses Durable Objects and Cloudflare Containers.
Can I use Workers AI for my coding agents?
Workers AI powers platform features, not the coding agents themselves. Agents use their own models (Claude, GPT, Gemini, Devstral) via their respective API keys.
What models does SAM use?
Llama 4 Scout 17B for general inference, GLM-4.7 Flash for task titles, Gemma 4 26B for context summarization, Qwen 3 30B for complex tasks, and Deepgram Aura 2 for text-to-speech.
Start running Workers AI on your infrastructure
Self-host on low-cost Cloudflare Workers. Bring your own cloud. Your agents, your code.