Claude Code changed how I write software. But it’s not the only CLI-based agentic coding tool out there. Here’s what else exists.
This list will age. For a live pulse check, OpenRouter keeps a coding apps ranking based on token usage. It’s a useful way to spot tools that are suddenly getting traction.
OpenAI Codex CLI
OpenAI’s answer to Claude Code. Install via npm i -g @openai/codex or brew install --cask codex. Works the same way—reads your repo, makes edits, runs commands.
Codex recommends GPT-5.5 when it’s available, with GPT-5.4 and GPT-5.3-Codex as fallback options. Peter Steinberger says Codex sometimes reads files for 10-15 minutes before writing, which helps it beat Claude Opus on larger tasks.
Included with ChatGPT Plus, Pro, Business, Enterprise, and Edu subscriptions.
Amp
Originally built at Sourcegraph, now an independent company. Available as CLI and in VS Code, Cursor, Windsurf, JetBrains, Neovim, and Zed. Mitchell Hashimoto (HashiCorp co-founder) uses Amp for his work on Ghostty.
Three main modes: “smart” (Opus 4.7), “rush” (faster, cheaper for small tasks), and “deep” (GPT-5.5 for extended reasoning). Has an agentic code review tool and MCP support with lazy-loading.
Pricing: Amp Free exists but may be waitlisted or full. Paid use is pay-as-you-go credits; individuals and non-enterprise workspaces pay provider API cost with no markup.
Factory
An AI-native development platform built around its CLI agent, Droid. Install and run it in the terminal like Claude Code, but Factory goes further—Droids also run via IDE, Slack, Teams, browser, or triggered automatically from issue trackers.
What stands out: “Droid Computers” give each agent a persistent cloud machine for remote work. Supports multi-step missions, code review, and non-interactive CI/CD execution. Vendor-agnostic with multiple model providers. Configuration uses AGENTS.md files and custom skills.
Backed by Khosla Ventures and Sequoia Capital ($150M Series C, $1.5B valuation). Used by Podium, Groq, and Chainguard. Pricing is no longer enterprise-only: Factory lists Pro, Plus, and Max individual plans, with Teams and Enterprise tiers above them.
OpenCode
Open source with 150,000+ GitHub stars. Supports 75+ LLM providers through Models.dev—Claude, GPT, Gemini, local models, OpenRouter, and OpenAI-compatible endpoints. Available across terminal, desktop apps (macOS, Windows, Linux), and IDE extensions.
You can share session links for collaboration. It supports ChatGPT Plus/Pro and GitHub Copilot logins; Claude access is safest through Anthropic API keys.
Aider
The OG terminal-first AI pair programmer. 44,000+ GitHub stars, 6.8M pip installs. Works with Claude, GPT, DeepSeek, and local LLMs.
What makes it different: Git-native workflow. Aider automatically commits changes with descriptive messages. It maps your entire codebase for context and supports 100+ programming languages. Also has voice-to-code if you’re into that.
Gemini CLI
Google’s open source option (Apache 2.0). Free tier gives you 60 requests/min and 1,000 requests/day with a personal Google account, plus access to Gemini 3 with a 1M token context window.
Uses a ReAct loop with built-in tools for Google Search, file operations, shell commands, and web fetching. MCP support included.
Armin Ronacher (creator of Flask) uses Gemini CLI inside Claude Code for certain tasks.
Pi
Less well-known but worth mentioning. Created by Mario Zechner. Pi is a minimal terminal coding harness inside a broader agent toolkit, not a sealed product.
It supports 15+ providers, including Anthropic, OpenAI, Google, Bedrock, Mistral, Groq, Cerebras, xAI, Hugging Face, OpenRouter, Ollama, and more. The monorepo also includes an agent runtime, a unified LLM API, a TUI library, and web UI components. Slack and chat automation now live in the separate pi-chat repo.
Armin Ronacher builds a game from scratch with Opus and Pi in one of his videos.
The pattern: Claude Code is the default choice, but developers who want model flexibility gravitate toward OpenCode, Aider, or Pi. Those deep in the OpenAI ecosystem use Codex. Google loyalists get a surprisingly capable free tier with Gemini CLI.
Most power users I’ve seen run multiple tools. Different models excel at different tasks. It’s worth experimenting with a few of these to find one that fits how you work—your communication style, the languages you use, and the frameworks you build with all affect which tool clicks. Someone doing Swift/iOS might have a different experience than someone building Svelte web apps.