Setup Agent

AI Agent responsible for installing all the necessary tools and dependencies and running the app, built with Python and LangGraph.

Getting started

Installation

pipx

The easiest way to get started with the setup-agent is to use pipx for installing the agent as the CLI tool.

brew install pipx
pipx install setup-agent
setup-agent setup run

Development environment

1. Install uv according to the Astral docs
2. In your terminal run the following commands:

uv sync # install all dependencies
uv run pre-commit install # install pre-commit hooks

For the testing purposes, run:

source .venv/bin/activate
python src/cli/app.py setup run

CLI

In order to install the CLI globally, use provided installation scripts:

uv build # or ./scripts/build.sh
./scripts/install.sh

Then you can use the tool globally by running:

setup-agent setup run

In order to uninstall the tool globally:

./scripts/uninstall.sh

Usage

Environment variables

To make the tool working you will need 2 API keys:

• TAVILY_API_KEY - key for Tavily (web search layer for agents)
• an API key for your LLM provider, e.g OPENAI_API_KEY for OpenAI models (i.e. gpt-4o) or ANTHROPIC_API_KEY for Anthropic models (i.e. claude-sonnet-4-5). For full list of possible integrations, please check out these Langchain docs (and make sure to install langchain package for given provider using uv if you're using provider other than OpenAI or Anthropic).

The easiest way to use the tool is to add these 2 API keys to the .env file located where the cli is used:

# .env
TAVILY_API_KEY=...
OPENAI_API_KEY=...
ANTHROPIC_API_KEY=...

For extended debugging and tool usage metrics, please consider adding these Langsmith related variables:

# .env
LANGSMITH_TRACING=...
LANGSMITH_ENDPOINT=...
LANGSMITH_API_KEY=...
LANGSMITH_PROJECT=...

Configuration options

The CLI runs an interactive script that allows user to configure everything and define the problem:

Option	Type	Description	Default	Required
Project root	str	Path to the root of the project (needs to be a valid directory)	Current directory	Yes
Guideline files	List[str]	Optional list of guideline files. If no guidelines files are specified at this point, agent will suggest all relevant files to the user.	-	No
Task	str	Predefined goal to achieve by the agent. If task is not defined, agent will suggest some tasks to the user.	-	No
Model	str	LLM model to be used as a core reasoning model. For a full list of available models check out these Langchain docs	anthropic:claude-sonnet-4-5	No

The script allows for optional, extended model configuration as well:

Option	Type	Description	Default	Required
Temperature	float	Controls randomness of the model. Must be a number between 0.0 and 1.0	Model's default value	No
Max output tokens	int	Max output tokens of the model.	Model's default value	No
Timeout	int	Timeout for LLM calls in seconds.	Model's default value	No
Max retries	int	Max number of retries for LLM calls.	Model's default value	No