Ollama, a tool for running large language models (LLMs) locally, has released a new command called ollama launch that sets up and starts coding tools with a single command[1]. It supports four tools — Claude Code, OpenCode, Codex, and Droid — and lets you use them right away with local or cloud models, without preparing any environment variables or configuration files[1].

What "ollama launch" Is — A Launcher That Needs No Config Files

ollama launch is a new command for setting up popular coding tools such as Claude Code, OpenCode, and Codex together with local or cloud models, and then launching them[1]. Its biggest feature is that you do not need to prepare environment variables or configuration files yourself[1].

When you try to connect a model to a coding tool from the terminal, you often have to set the connection address and authentication token manually as environment variables. ollama launch bundles this preparation into a single command, guiding you through model selection and launch[1].

How to Use It — Installing Ollama and Launching in One Command

First, install Ollama v0.15 or later[1]. Then, in the terminal, pull the model you want to use. If you are using a local model that runs on your own machine, specify it like this[1]:

# Running with a 64000-token context length requires about 23 GB of VRAM
ollama pull glm-4.7-flash

When you need a larger context length, you can also pull a cloud model[1]:

ollama pull glm-4.7:cloud

Once the model is ready, specify the tool you want and launch it. For Claude Code, the command is as follows[1]:

ollama launch claude

For OpenCode, it is as follows[1]:

ollama launch opencode

Running the command walks you through selecting a model and launching your chosen tool. Here, too, there is no need to create environment variables or configuration files[1]. If you want to finish only the configuration without launching right away, add "--config" at the end[1]:

ollama launch opencode --config

Supported Coding Tools and Recommended Models

The tools ollama launch supports are Anthropic's coding tool Claude Code, along with OpenCode, Codex, and Droid — four tools in total[1]. All of them run in the terminal as developer-facing coding tools and can be connected to the models Ollama handles.

Ollama also lists recommended models for coding use[1]. For local models that run on your own machine, it names glm-4.7-flash, qwen3-coder, and gpt-oss:20b, while for cloud models it lists glm-4.7:cloud, minimax-m2.1:cloud, gpt-oss:120b-cloud, and qwen3-coder:480b-cloud[1]. You can choose between local and cloud depending on your machine's performance and the size of the model you want to use.

Context Length and the 5-Hour Coding Session

For work that handles long context, such as coding, Ollama recommends setting the context length (the number of tokens a model can handle at once) to at least 64000 tokens[1]. You can change this setting from Ollama's settings screen[1].

For cases where running these models locally is difficult, Ollama also offers a cloud service that provides hosted models[1]. This cloud supports the full context length and is said to come with relatively generous usage limits even on the free tier[1]. Along with this update, the usage allowance was expanded and an extended 5-hour coding session window was added[1]. Pricing details can be checked on Ollama's pricing page[1].

Summary

ollama launch is a new command that lets you set up and start coding tools such as Claude Code, OpenCode, Codex, and Droid from a single command, without environment variables or configuration files[1]. It is available in Ollama v0.15 and later and works with both local and cloud models[1]. For coding use, a context length of at least 64000 tokens is recommended, and when running locally is difficult, a cloud service that supports the full context length is also available[1]. For anyone who wants to try coding tools on their own hardware or with open models, it can be called a change that lowers the barrier to getting started.

Source: https://ollama.com/blog/launch