Sub-50ms Time to First Token
Newton’s runs on Cloudflare Workers deployed across 330+ edge locations worldwide. Your requests terminate at the nearest regional data center, cutting network round trips to near zero.
RUST-SPEED EDITOR INTEGRATION
Zed is engineered in Rust for millisecond-level responsiveness. Don't bottle up your editor with sluggish AI gateways. Route Zed through Newton’s global Cloudflare edge for sub-50ms TTFT, prompt caching savings, and open model freedom.
Why modern developers connect Zed’s Assistant to Newton’s edge proxy.
Newton’s runs on Cloudflare Workers deployed across 330+ edge locations worldwide. Your requests terminate at the nearest regional data center, cutting network round trips to near zero.
Zed supports custom OpenAI endpoints natively in ~/.config/zed/settings.json. Simply set api_url to Newton’s and launch your assistant.
Skip $20/month fixed subscriptions for seats you only use part-time. With Newton’s, you only pay for the tokens you stream, with zero idle fees and full prompt caching discounts.
Add this to your ~/.config/zed/settings.json file.
{
"language_models": {
"openai": {
"version": "1",
"api_url": "https://newtons.dev/api/v1",
"available_models": [
{ "name": "gpt-5.6-sol", "max_tokens": 128000 },
{ "name": "claude-sonnet-5", "max_tokens": 64000 },
{ "name": "deepseek-reasoner", "max_tokens": 64000 }
]
}
}
}Export your key so Zed picks it up automatically:
export OPENAI_API_KEY="newt_your_key_here"Open the Assistant panel with cmd-? (macOS) or ctrl-? (Linux) and enjoy blazing fast token streaming.
See how Zed's rapid buffer streaming exhausts direct API tier token caps and how Newton's pooled failover and prompt caching keeps you coding uninterrupted.
Read Zed 429 Troubleshooting Guide →