Aider

Aider is an open-source AI pair programming tool for the terminal. It integrates tightly with Git, committing changes automatically as you work with it. With Claude Sonnet as the backend, it scores ~72% on SWE-Bench Verified as of early 2026.

Why It's in Trial

Aider has been around longer than most agentic coding tools and has earned a reputation for reliability and a thoughtful workflow:

• Git-native: Every change Aider makes is committed to Git with a descriptive message, giving you a full history of what the AI did and easy rollback.
• Model agnostic: Works with Claude 3.7 Sonnet, GPT-4o, o1/o3-mini, DeepSeek R1/V3, Gemini, and local models via Ollama. DeepSeek has become popular for cost-effective use.
• Architect + Editor mode: A two-model workflow where an "architect" model (e.g., DeepSeek R1) plans changes and an "editor" model (e.g., Claude Sonnet) implements them — a major capability for complex tasks.
• Diff-based editing: Aider shows you exactly what it's changing before it changes it.
• Multi-file awareness: Add any files to the context and Aider can edit across all of them, with an internal repo map for large codebases.
• Automated linting and testing: Aider can run linters and tests after each change and self-correct on failures.

Getting Started

# Recommended: install script (no Python required)
curl https://aider.chat/install.sh | sh

# Or via pip installer
pip install aider-install && aider-install

# Use with Claude (requires ANTHROPIC_API_KEY)
aider --model sonnet

# Or with GPT-4o (requires OPENAI_API_KEY)
aider --model gpt-4o

# Add files to context
/add src/main.py tests/test_main.py

Compared to Claude Code

Both are terminal-based agentic tools. Aider's Git integration, open-source nature, and model flexibility are strengths; Claude Code's deeper integration with Anthropic's models and its interactive permission model feel more polished for many workflows.

Key Characteristics