Zed Editor setup with Ollama and LM Studio

This tutorial connects Zed Editor to a local LLM through Ollama or LM Studio, then shows how to use a model running on another computer in the same network. The speaker mentions Zed version 1.0 and demonstrates switching between models on a laptop and a desktop with a stronger GPU.

For each runner, the speaker first checks that a model responds outside Zed. In the editor's agent panel, the LLM provider settings contain default local URLs for both services. Connecting retrieves the available models. The Ollama example uses Gemma 4 and raises the context window from 8,000 to 60,000 tokens because the speaker considers the smaller setting insufficient for tool use. This is a demonstration setting, rather than a general requirement.

The guide also explains how an agent profile controls permission to create or edit project files. A thinking switch enables or disables reasoning when the selected model supports it; the speaker suggests experimenting with it if responses are slow.

For network access, the speaker adds an OpenAI-compatible provider to keep both local and desktop connections available. The form requires a default model name. The speaker says Ollama needs no API key by default and enters arbitrary text in that field. An initial connection fails until the API URL includes the full path, port and protocol version. Remote LM Studio access is described as the same process with a different URL, but is not demonstrated.