Apps · IDEs & editors

Zed

High-performance collaborative editor with an agent panel.

Tokens / 7d

63B

Routing directive

auto:coding

On our metal

12%

Licence

Open source

Zed's assistant panel and inline transforms are built for speed. Requests are short-lived and heavily parallel, which rewards providers with stable tail latency rather than the best average.

→ zed.dev → All apps → Integration docs

How it connects

Custom OpenAI-compatible provider in settings.json.
Streaming inline transforms with aggressive cancellation.
P99 latency routing profile enabled.

Traffic pattern

Latency-sensitive and cancellation-heavy; a third of streams are aborted before completion.

claude-sonnet-4.5gpt-5.5gemini-3.6-flash

Point it at the gateway

// Zed settings.json
"language_models": { "openai": {
  "api_url": "https://api.llmcloud.ai/v1",
  "available_models": [{ "name": "auto:coding", "max_tokens": 200000 }]
}}