At home in your terminal.
Read, search, edit, and run commands in your checkout. Turn a conversation into changes you can inspect.
A local coding agent for Linux
Your code.
Your machine.
Your Echo.
A local coding agent for your terminal. Explore a repository, make a change, and review the result—with models you run yourself.
Python 3.12+ · Linux · Local models
A question is a good place to start.
PythonvLLMLiteLLMDocker
From a question
to a working change.
Ask Echo to find its way through a repository, work on a feature, or show what changed. Keep the conversation next to the code.
Start your first sessionRead. Search. Edit. Run.
The tools you use, in one conversation.
echo / terminalIllustrative session
uv run echo-ai
How does session persistence work?
EchoSearching the repository
search src/echo_ai/
read session storage and runtime
trace how conversations are saved
Find the context before making a change.
Bring your own models.
Run Qwen or Gemma locally with vLLM. Echo connects through LiteLLM, with model settings you control.
Keep the thread.
Resume saved sessions and revisit recorded edits. Try a disposable Docker workspace with sandbox mode.
Room to experimentTry it in a
Try it in a
fresh workspace.
Use sandbox mode to work in a disposable Docker copy. Inspect recorded agent edits, then apply them back to your checkout when you’re ready.
Read about sandbox mode# Start in a disposable workspace
$ uv run echo-ai chat --sandbox
$ uv run echo-ai chat --sandbox
# Review recorded agent edits
/diff
/diff
# Bring them into your checkout
/apply
/apply
Shell changes are outside /diff and /apply tracking.
1
Clone the repository
git clone https://github.com/KaiMJ/echo-ai.git2
Install the dependencies
cd echo-ai && uv sync --lockedLinux · Python 3.12+ · uv · Git · Bash · ripgrep
Local inference requires Docker with NVIDIA GPU support.