OpenJet
Offline coding agent that sets up and runs local models on your own GPU.
Tested on Linux
Use it
Linux x86_64
You need first
- System packages:
CUDA toolkit - A GPU is optional
Install open-jet via uv tool
uv tool install open-jet
Check that it works
openjet --help 2>&1The output should contain “offline agentic terminal UI”.
First thing to try
openjet --status 2>&1Run openjet --status first to check your setup.
What it does
- Run a coding agent in the terminalNot shown working
- Run four coding agents at once on one GPUNot tested
- Plan with a frontier model and code with a local oneNot tested
- Embed a local model in a Python appNot shown working
How far pickfit got
Linux x86_64
- Installs
- Starts
- 0 of 4 features work
- Works offline
Good for
- run a coding agent in the terminal
- run 4 coding agents at once on one GPU
- have a frontier model plan while local model codes
- ship a local model inside a Python app
- read repo, edit files and run tests locally
An alternative to llama.cpp, Ollama and LM Studio.
How pickfit tested it
Each run starts from a clean machine and follows the steps above. How pickfit verifies
Linux x86_64
Its main function works
- Replayed from scratch
- Works offline
- Checked on 30 Oct 2026
- Version 0.5.4
Differences the test machine hit
- README: pipx install open-jet / pip install open-jet
In the test: uv tool install open-jet isolates the package; system python cannot import openjet - README: openjet setup downloads a model and configures the agent
In the test: on this small CPU-only box setup --dry-run exits 1 with 'No selected model fits available memory with runtime headroom'