shai is a coding agent, your pair programming buddy that lives in the terminal. Written in rust with love <3
Install the latest release with the following command:
curl -fsSL https://raw.githubusercontent.com/ovh/shai/main/install.sh | sh
the shai binary will be installed in $HOME/.local/bin
Install the last unstable version with the following command:
curl -fsSL https://raw.githubusercontent.com/ovh/shai/main/install.sh | SHAI_RELEASE=unstable shthe shai binary will be installed in $HOME/.local/bin
By default shai uses OVHcloud as an anonymous user meaning you will be rate limited! If you want to sign in with your account or select another provider, run:
shai auth
Once you have a provider set up, you can run shai:
shai
Shai can also run in headless mode without user interface. In that case simply pipe a prompt into shai, it will stream event in the stderr:
echo "make me a hello world in main.py" | shai
you can also instruct shai to return the entire conversation as a trace once it is done:
echo "make me a hello world in main.py" | shai 2>/dev/null --trace
this is handy because you can chain shai calls:
echo "make me a hello world in main.py" | shai --trace | shai "now run it!"
You can create a SHAI.md file at the root of your project containing any information you want Shai to know about the project (architecture, build steps, important directories, etc.). Shai will automatically load this file as additional context.
Instead of a single global configuration, you can create custom agent in a separate configuration.
example.config contains an example of a custom configuration with an stdio MCP server configured.
Place this file in ~/.config/shai/agents/example.config, you can then list the agents available with:
shai agent list
you can run shai with this specific agent with the agent subcommand:
shai example
shai can also act as a shell assistant in case a command failed and will propose you a fix. This works by injecting command hook while monitoring your terminal output. Your last terminal output along with the last command and error code will be sent for analysis to the llm provider. To start hooking your shell with shai simply type:
$ shai on
for instance:
To stop shai from monitoring your shell you can type:
$ shai off
Simply build the project with cargo
git clone git@github.com:ovh/shai.git
cd shai
cargo build --release
OVHCloud provides compatible LLM endpoints for using shai with tools. Start by creating a Public Cloud project in your OVHCloud account, then head to AI Endpoints and retreive your API key. After setting it in shai, you can:
- choose one of the models with function calling feature (e.g., gpt-oss-120b, gpt-oss-20b, Mistral-Small-3.2-24B-Instruct-2506) for best performance ;
- choose any other model forcing structured output (
/set sooption).
