Getting Started with Aider

Aider is an open-source AI pair-programming tool that lives in your terminal and works with your local git repository. You chat in the shell; Aider edits the files you add to the session, shows diffs, and can commit changes with messages it drafts. Official docs recommend installing into an isolated environment via aider-install, uv, or pipx, then running aider inside a project folder with a model and API key. Homepage and install guide: aider.chat and aider.chat/docs/install.html.

This guide is for after-work learners who already have Python and git on the machine, and want a terminal pair partner for small scripts—not a second full-time job configuring models. One evening is enough to install, open a practice repo, and complete one reviewable edit.

If you are surveying several agents, our Hub roundup on top AI agent tools for after-work Python sits alongside this piece. Tonight you only need Aider to feel calm and reversible.

Who this is for

• People who can run python / python3 and git in a terminal
• Learners who like seeing diffs and using /undo more than opaque IDE magic
• Anyone with 60–90 minutes and a provider API key they already pay for (or will set up carefully)

If git is still unfamiliar, initialise a tiny practice repo tonight and keep work repos closed.

Steps (one evening)

1. Install Aider the recommended way

Official quick start (Python 3.8–3.13 already installed):

python -m pip install aider-install
aider-install

That installs Aider in its own environment. Alternatives documented by Aider include uv tool install and pipx install aider-chat. Pick one. Do not stack three installers in the same sitting.

Confirm:

aider --help

2. Create a disposable git practice repo

mkdir -p ~/aider-practice && cd ~/aider-practice
git init
echo "# practice" > README.md
git add README.md && git commit -m "start practice"

Aider is git-native. A clean practice repo makes /undo and commits easy to understand. Keep .env files and work secrets out of this folder.

3. Launch with a model and key (without pasting keys into chat history)

From the project directory, follow Aider’s documented pattern—for example with an OpenAI-compatible key passed on the command line or via environment variables / a local .env that you never commit:

# Example shape from official docs — replace with your provider and key handling
aider --model o3-mini --api-key openai=<your-key>

Other documented examples use --model sonnet with an Anthropic key, or DeepSeek. Prefer env vars or a gitignored .env over pasting keys into screenshots. If the model name fails, check Aider’s current model docs—names change; the workflow does not.

4. Add only the files you need, then ask for one change

/add README.md

Or start with aider README.md. Ask for one concrete edit, for example: “Add a short section titled Evening goal with three bullet points.” Review the diff. Let it commit only if the message and patch look right. Use /undo if not.

5. End on a green commit

git log -1 --oneline
git status

Stop when one commit is good. Resist “also refactor everything.”

Tiny exercise (10 minutes)

Create notes.py (empty or a one-line comment). /add notes.py. Ask Aider to add a main() that prints today’s date using the standard library, then run python notes.py yourself. Open the file and rename one variable by hand.

Pitfalls

• Adding the whole repo to the chat. Aider’s own usage docs say to add only files that need editing. Extra files cost tokens and confuse the model.
• No git repo. Initialise git first; otherwise you lose the commit//undo safety net.
• Committing secrets. Never /add .env or credential files. Confirm .gitignore before the first commit that is not a toy.
• Treating auto-commits as review. Read the diff every time. /undo exists for a reason.
• Chasing every model brand in one night. One working provider is enough for learning the loop.

Next step

Aider speeds edits; it does not replace learning Python syntax and structure. When you want short, after-work lessons—not a three-day workshop—start with Introduction to Python Programming.