Key takeaways
- 46K+ GitHub stars and a beloved community, but development has slowed to maintenance mode — last feature release was v0.86.0 in August 2025
- Model-agnostic with 75+ providers — works with Claude, GPT, DeepSeek, Gemini, and local models, though official model recommendations are now stale
- Automatic git integration — commits changes with sensible messages, making AI changes easy to track and undo
FAQ
What is Aider?
Aider is an open-source AI pair programming tool that runs in your terminal, letting you have natural conversations with LLMs to edit code in your local git repository.
Is Aider free?
Yes, Aider is free and open source (Apache 2.0). You provide your own API keys or use local models.
What models work with Aider?
Aider supports 75+ LLM providers including local models via Ollama, so current frontier models (Claude Opus 4.x and Fable 5, GPT-5.x, Gemini 3) work through its provider layer. Note that Aider's official docs still recommend older models like Claude 3.7 Sonnet, DeepSeek R1/V3, and OpenAI o1/o3-mini — the guidance hasn't kept pace with model releases.
Is Aider still maintained?
Partially. The repository still receives commits as of May 2026, but the last feature release was v0.86.0 in August 2025, with only a patch release (v0.86.2) in February 2026. Aider is best described as being in maintenance mode, not abandoned.
How is Aider different from Claude Code?
Aider is model-agnostic and community-driven; Claude Code is Anthropic-only but more deeply optimized for Claude models and now iterates much faster than Aider.
Executive Summary
Status note (June 2026): Aider is alive but in maintenance mode. The repository still receives commits (most recently May 22, 2026, adding new Claude model names), but the last feature release was v0.86.0 in August 2025,[1] with only a patch release (v0.86.2, February 2026) since.[2] The LLM Leaderboards were last updated November 20, 2025.[3]
Aider is an open-source AI pair programming tool that defined the terminal-based coding agent category. With roughly 46,000 GitHub stars as of June 2026 and a passionate community, it has frequently been cited as "the best open-source AI coding assistant" by developers.[4] Created by Paul Gauthier (Aider AI LLC), Aider pioneered many patterns now common in AI coding tools — automatic git commits, repository mapping, and architect/editor workflows. Its development pace, however, has slowed markedly since mid-2025 while competitors like Claude Code ship weekly.
| Attribute | Value |
|---|---|
| Creator | Paul Gauthier / Aider AI LLC |
| Founded | 2023 |
| Funding | Bootstrapped |
| GitHub Stars | ~46,000 (as of June 2026)[4] |
| License | Apache 2.0 |
| Last Feature Release | v0.86.0 (August 2025)[1] |
Product Overview
Aider lets you pair program with LLMs to start a new project or build on your existing codebase.[5] You run it in your terminal, describe what you want to change, and Aider edits your files directly — automatically committing changes to git with sensible commit messages.
Unlike IDE-based tools that add AI to your editor, Aider takes the opposite approach: it's a terminal-first tool that works alongside any editor. You can run Aider in one terminal while editing in VS Code, Vim, or whatever you prefer.
Key Capabilities
| Capability | Description |
|---|---|
| Natural Language Coding | Describe changes in plain English; Aider edits files |
| Repository Mapping | Analyzes codebase structure to help with large projects |
| Automatic Git Integration | Commits every change with descriptive messages |
| Voice-to-Code | Speak your requests using built-in voice input |
| Multi-Model Support | Works with 75+ LLM providers |
Product Surfaces
| Surface | Description | Availability |
|---|---|---|
| Terminal CLI | Primary interface (aider command) | GA |
| IDE Watch Mode | Add comments in your editor, Aider implements | GA |
| Docker | Containerized deployment option | GA |
Technical Architecture
Aider installs via pip or uv and runs entirely locally.[6] It doesn't require any server infrastructure — you just need API keys for your chosen LLM provider.
Installation:
# Quick install
python -m pip install aider-install
aider-install
# Start with your preferred model
aider --model sonnet --api-key anthropic=<key>
aider --model deepseek --api-key deepseek=<key>
aider --model o3-mini --api-key openai=<key>
Note that the official docs' model examples (Sonnet aliases, o3-mini, DeepSeek) reflect the early-2025 model generation; newer frontier models work via explicit model strings through Aider's LiteLLM-based provider layer.
Key Technical Details
| Aspect | Detail |
|---|---|
| Deployment | Local CLI (no server required) |
| Model(s) | 75+ providers — Anthropic, OpenAI, Google, DeepSeek, local |
| Integrations | Git, any text editor (via watch mode) |
| Open Source | Yes (Apache 2.0) |
Repository mapping: Aider creates a compact map of your entire codebase, helping the LLM understand project structure even when only a few files are in the chat.[4]
Architect mode: For complex tasks, Aider can use a "reasoning" model to plan changes and a "coding" model to implement them, combining their strengths.
Strengths
- Influential benchmarks — Aider's LLM Leaderboards were long the de facto standard for evaluating coding model performance, though they were last updated November 2025[3]
- Git-native workflow — Every AI change is a clean commit; easy to review, revert, or iterate
- Model flexibility — Swap providers as new models release; no vendor lock-in
- Beloved by developers — "It's going to rock your world" (Eric S. Raymond); "Quadrupled my coding productivity" (HN users)
- Low resource usage — Terminal-based, minimal memory footprint
- 100+ language support — Works with Python, JavaScript, Rust, Go, and dozens more
- Mature and stable — Years of refinement; the core edit-commit loop is battle-tested
Cautions
- Development has slowed sharply — No feature release since v0.86.0 in August 2025; only a patch (v0.86.2) in February 2026[1][2] — while Claude Code, Codex CLI, and Gemini CLI ship continuously
- Stale model guidance — The official site still recommends Claude 3.7 Sonnet, DeepSeek R1/V3, and o1/o3-mini as of June 2026[5] — a model generation behind current frontier options
- Leaderboard no longer current — Last updated November 20, 2025; the current model generation is unbenchmarked there[3]
- Terminal-only — No GUI; requires comfort with command line
- No background execution — Runs in foreground, requires active attention; lacks the subagent/background-task features newer agents have added
- BYOK required — No free tier; users must provide their own API keys
- Limited IDE integration — Watch mode works but isn't as seamless as native extensions
- Community support only — No enterprise SLAs or dedicated support
Pricing & Licensing
| Tier | Price | Includes |
|---|---|---|
| Aider | Free | Open source tool (Apache 2.0) |
| LLM API | Varies | Pay your chosen provider (Anthropic, OpenAI, etc.) |
Licensing model: Free and open source — you only pay for LLM API usage.
Hidden costs: Heavy usage with frontier models (Claude Opus/Sonnet 4.x, GPT-5.x) can cost $20-100+/month depending on usage intensity. DeepSeek and other budget providers offer much lower per-token costs. Unlike Claude Code or Codex CLI, Aider has no subscription bundle option — API metering only.
Competitive Positioning
Direct Competitors
| Competitor | Differentiation |
|---|---|
| Claude Code | Claude Code is Anthropic-native with rapid release cadence and agentic features (subagents, background tasks); Aider is model-agnostic but iterating far more slowly |
| OpenCode | Both are open source terminal agents; OpenCode is actively developed with multi-provider support and has absorbed much of the open-source CLI mindshare Aider once held |
| Cline | Cline is VS Code-native with GUI approvals; Aider is terminal-native |
| Codex CLI / Gemini CLI | First-party open-source CLIs from OpenAI and Google, both shipping frequently; Aider competes as the neutral, provider-agnostic incumbent |
| Cursor | Cursor is a full IDE; Aider is a CLI tool that works with any editor |
When to Choose Aider Over Alternatives
- Choose Aider when: You live in the terminal, want model flexibility, value git-native workflows, and prefer a stable, mature tool over a fast-moving one
- Choose Claude Code when: You're committed to Claude models and want Anthropic's official, rapidly evolving tooling
- Choose OpenCode when: You want an actively developed open-source terminal agent with multi-provider support
- Choose Cline when: You prefer VS Code with a visual interface for approvals
- Choose Cursor when: You want a full AI-native IDE experience
Ideal Customer Profile
Best fit:
- Developers who live in the terminal
- Users who want to try different LLM providers
- Teams that value clean git history with traceable AI changes
- Individual developers who prefer lightweight tools over heavy IDEs
- Projects needing support for multiple programming languages
Poor fit:
- Teams requiring enterprise support or SLAs
- Developers uncomfortable with command-line interfaces
- Organizations needing GUI-based approval workflows
- Users who want "set and forget" background agents
Viability Assessment
| Factor | Assessment |
|---|---|
| Financial Health | Stable — Bootstrapped, no burn; sustainability depends on maintainer attention |
| Market Position | Eroding leader — Still the most-starred open source coding CLI, but mindshare is shifting to Claude Code, Codex CLI, and OpenCode |
| Innovation Pace | Slowed — No feature release since August 2025; commits continue but are mostly model-list updates and patches[1] |
| Community/Ecosystem | Large but cooling — ~46K stars as of June 2026,[4] active Discord[7] |
| Long-term Outlook | Uncertain — Category-defining tool, but at risk of being lapped by first-party CLIs unless development re-accelerates |
Aider achieved the rare status of being both the most popular and the most respected tool in its category, and the LLM Leaderboards were essential infrastructure for the AI coding ecosystem through 2025. As of June 2026, however, the project meets the bar for "maintained" (commits within the past 30 days) but not "actively developed" — the leaderboards have not benchmarked the current model generation, and the docs recommend models a generation old.[5]
Bottom Line
Aider remains a capable, stable terminal-based coding assistant with the cleanest git-native workflow in the category, and its model-agnostic approach means no vendor lock-in. But it is no longer the pacesetter: development has been in maintenance mode since mid-2025, while first-party CLIs from Anthropic, OpenAI, and Google iterate weekly.
Recommended for: Developers who live in the terminal, want provider flexibility with clean git integration, and are comfortable with a mature tool that changes slowly.
Not recommended for: Teams needing enterprise support, GUI approval workflows, the latest agentic features (subagents, background execution), or anyone betting on rapid future development.
Outlook: Aider's legacy as the reference implementation for terminal coding agents is secure, but its trajectory is flat. Unless release cadence recovers, expect it to settle into "stable classic" status while newer agents define the frontier. Watch for whether the leaderboards resume updates — they were the project's clearest signal of vitality.
Research by Ry Walker Research • methodology