What is a local LLM coding agent?

It is an agent that runs its language model locally and can use controlled tools to inspect, edit, and test code in a repository.

How is this different from local LLM chat?

Chat only returns text. A coding agent also needs file tools, shell permissions, session state, repository context, and a reliable tool loop.

Which local model does OpenJet use?

Setup chooses from a hardware-aware catalog. The current documented options include Qwen3.8-27B and smaller Qwen models for lower memory budgets.

local LLM coding agent

Run a local LLM as a real coding agent

A local chat window can answer questions. A local coding agent needs tools, state, permissions, and a model configuration that remains usable across a real repository task.

OpenJet turns an open model into a repository-aware terminal agent. It handles model selection, runtime setup, file tools, approved shell execution, context condensing, and resumable sessions without a hosted inference API.

Install OpenJet Read the docs

Why local wins

Works across a repository. Read, grep, glob, create, and edit tools let the model operate on actual project files instead of relying on pasted snippets.
Configures the model layer. Setup detects RAM and GPU memory, selects a fitting profile, downloads the model, and configures llama.cpp.
Keeps execution reviewable. The model can propose shell commands, but OpenJet requires approval before state-changing execution.
Survives longer tasks. Resumable state and context condensing are designed for local models with tighter context budgets.

Common deployment paths

Fix a bug across several local files without uploading the repository.
Use Qwen3.8-27B as a terminal coding agent on a 20 GB profile.
Review code in a sensitive or disconnected environment.
Benchmark model and runtime settings on the actual development machine.

Frequently asked questions

What is a local LLM coding agent?: It is an agent that runs its language model locally and can use controlled tools to inspect, edit, and test code in a repository.
How is this different from local LLM chat?: Chat only returns text. A coding agent also needs file tools, shell permissions, session state, repository context, and a reliable tool loop.
Which local model does OpenJet use?: Setup chooses from a hardware-aware catalog. The current documented options include Qwen3.8-27B and smaller Qwen models for lower memory budgets.

Explore OpenJet

Embed a local model with the Python SDK, run Qwen3.8-27B as a local coding agent, compare local coding-agent alternatives, read the docs, or go back to the homepage.