# Best open-source AI repos for AI coding agents (2026) Source: https://dedupedai.com/best-ai-repos-for/ai-coding-agents/ (Deduped, updated 2026-10-27). Markdown twin of the page; JSON: https://dedupedai.com/best-ai-repos-for/ai-coding-agents/index.json ## What is the best open-source AI coding agent? The best open-source AI repos for AI coding agents right now are Claude Code, OpenCode and Gemini CLI. All three run in your terminal and are the most starred coding agents we track; Cline also works inside your IDE, and goose runs as a desktop app, a CLI and an API. ## Comparison | Repo | What it does | Best for | Stars | Language | |---|---|---|---|---| | [Claude Code](https://dedupedai.com/use-cases/anthropics-claude-code/) (anthropics/claude-code) | Claude Code is an agentic coding tool that runs in your terminal, reads your codebase, and carries out coding and git tasks from natural language commands. | Developers who work in a terminal and want an AI agent to explain code, run routine tasks, and handle git workflows inside a project. | 152,856 | TypeScript | | [OpenCode](https://dedupedai.com/use-cases/anomalyco-opencode/) (anomalyco/opencode) | OpenCode is an open source AI coding agent that runs as a command-line tool and as a beta desktop app. | Software developers and engineering teams who want an AI coding agent they can inspect, run locally, and switch between read-only and full-access modes. | 219,977 | TypeScript | | [Gemini CLI](https://dedupedai.com/use-cases/google-gemini-gemini-cli/) (google-gemini/gemini-cli) | Gemini CLI is Google's open-source AI agent for the terminal. | Developers with a Gemini API key or Google Cloud project who want a scriptable terminal agent. | 108,198 | TypeScript | | [Cline](https://dedupedai.com/use-cases/cline-cline/) (cline/cline) | Cline is an open source autonomous coding agent that runs as an SDK, an IDE extension, a terminal CLI, or a desktop app. | Developers and engineering teams who want an AI agent that edits code, runs commands, and automates recurring coding tasks. | 70,700 | TypeScript | | [Crush](https://dedupedai.com/use-cases/charmbracelet-crush/) (charmbracelet/crush) | Crush is Charm's agentic coding tool for the terminal. | Developers who live in the terminal and want a polished coding agent that works with the model provider of their choice, including local models. | 28,821 | Go | | [Goose](https://dedupedai.com/use-cases/aaif-goose-goose/) (aaif-goose/goose) | goose is a general-purpose, open-source AI agent that runs on your machine as a desktop app for macOS, Linux and Windows, a CLI and an API. | Developers and knowledge workers who want one local agent for code, research, writing and automation with the model provider of their choice. | 55,769 | Rust | ## When to pick which ### Claude Code (anthropics/claude-code) Claude Code is an agentic coding tool that runs in your terminal, reads your codebase, and carries out coding and git tasks from natural language commands. Best for: Developers who work in a terminal and want an AI agent to explain code, run routine tasks, and handle git workflows inside a project. Needs: A terminal on macOS, Linux, or Windows; A local project directory you want Claude Code to work in; Permission to run one of the install methods listed in the README. Start here: [Explain complex code in an unfamiliar project](https://dedupedai.com/use-cases/anthropics-claude-code/explain-unfamiliar-codebase/) (Intermediate, 15 minutes) 1. Install Claude Code with one of the README methods, for example curl -fsSL https://claude.ai/install.sh | bash on macOS or Linux. 2. Navigate to your project directory. 3. Run claude. 4. Ask in natural language to explain how a named module or feature works, and to list the files it depends on. All 3 use cases: https://dedupedai.com/use-cases/anthropics-claude-code/ | GitHub: https://github.com/anthropics/claude-code ### OpenCode (anomalyco/opencode) OpenCode is an open source AI coding agent that runs as a command-line tool and as a beta desktop app. Best for: Software developers and engineering teams who want an AI coding agent they can inspect, run locally, and switch between read-only and full-access modes. Needs: A terminal on macOS, Linux, or Windows, or the desktop app installer for your platform; One package manager available: npm, bun, pnpm, yarn, or brew; A local project folder you want the agent to work in; Configuration for OpenCode as described in the docs at opencode.ai/docs (the README only links to them). Start here: [Explore an unfamiliar codebase without risking edits](https://dedupedai.com/use-cases/anomalyco-opencode/explore-unfamiliar-codebase-read-only/) (Intermediate, 30 minutes) 1. Run npm i -g @opencode/cli@latest (or use brew install anomalyco/tap/opencode-v2 on macOS and Linux). 2. Change into the project folder and start OpenCode with opencode. 3. Press the Tab key until the plan agent is selected. 4. Ask it to map the main modules, entry points, and how the tests are run. All 5 use cases: https://dedupedai.com/use-cases/anomalyco-opencode/ | GitHub: https://github.com/anomalyco/opencode ### Gemini CLI (google-gemini/gemini-cli) Gemini CLI is Google's open-source AI agent for the terminal. It reads your codebase, runs commands and answers prompts with Gemini models, and can run headless with JSON or streaming JSON output for scripts and CI. The README notes that unpaid tier and Google One users were moved to Antigravity CLI on June 18, 2026 and that Gemini CLI is not optimized for the newest Gemini models, so it now fits API key and Vertex AI users best. Best for: Developers with a Gemini API key or Google Cloud project who want a scriptable terminal agent. Needs: Node.js (npm), Homebrew, MacPorts or conda; A Gemini API key from AI Studio or a Google Cloud project for Vertex AI. Start here: [Ask Gemini to explain a codebase from the terminal](https://dedupedai.com/use-cases/google-gemini-gemini-cli/ask-gemini-about-your-codebase-from-the-terminal/) (Beginner, 10 minutes) 1. Try it without installing: npx @google/gemini-cli, or install with npm install -g @google/gemini-cli or brew install gemini-cli 2. Set your key: export GEMINI_API_KEY="YOUR_API_KEY" 3. In the repo run gemini, and add sibling folders with gemini --include-directories ../lib,../docs 4. Ask it to explain the architecture or summarize recent changes All 3 use cases: https://dedupedai.com/use-cases/google-gemini-gemini-cli/ | GitHub: https://github.com/google-gemini/gemini-cli ### Cline (cline/cline) Cline is an open source autonomous coding agent that runs as an SDK, an IDE extension, a terminal CLI, or a desktop app. Best for: Developers and engineering teams who want an AI agent that edits code, runs commands, and automates recurring coding tasks. Needs: Node.js with npm available, since the README installs the CLI with npm; Access to an AI model provider that Cline supports, such as Anthropic, OpenAI, Google, OpenRouter, or a local model through Ollama or LM Studio; A git repository or project folder you want the agent to work in; For the VS Code path: the VS Code editor and the VS Code Marketplace extension. Start here: [Fix failing tests automatically in a CI pipeline](https://dedupedai.com/use-cases/cline-cline/fix-failing-tests-in-ci/) (Intermediate, 30 minutes) 1. Install the CLI with npm i -g cline. 2. Open a terminal in the root of the repository that has the failing tests. 3. Run cline "Run tests and fix any failures". 4. Review the file changes with git diff before committing. All 5 use cases: https://dedupedai.com/use-cases/cline-cline/ | GitHub: https://github.com/cline/cline ### Crush (charmbracelet/crush) Crush is Charm's agentic coding tool for the terminal. It works with many LLMs through OpenAI- and Anthropic-compatible APIs, lets you switch models mid-session while keeping context, keeps multiple sessions per project, uses LSPs for extra context and adds capabilities through MCP servers. It runs on macOS, Linux, Windows, Android and the BSDs and is configured with a Bash-like crushrc. Best for: Developers who live in the terminal and want a polished coding agent that works with the model provider of their choice, including local models. Needs: A terminal on macOS, Linux, Windows, Android or BSD; An LLM provider: Charm's Hyper subscription, an API key for Anthropic, OpenAI, Gemini or OpenRouter, or a local server such as Ollama. Start here: [Start an AI coding session in your terminal with Crush](https://dedupedai.com/use-cases/charmbracelet-crush/start-coding-with-crush-in-your-terminal/) (Beginner, 10 minutes) 1. Install with brew install charmbracelet/tap/crush or npm install -g @charmland/crush 2. On Windows use winget install charmbracelet.crush 3. Start Crush in your project, press ctrl+l to open the model picker and choose a provider 4. Paste your API key, or set a provider environment variable such as ANTHROPIC_API_KEY All 3 use cases: https://dedupedai.com/use-cases/charmbracelet-crush/ | GitHub: https://github.com/charmbracelet/crush ### Goose (aaif-goose/goose) goose is a general-purpose, open-source AI agent that runs on your machine as a desktop app for macOS, Linux and Windows, a CLI and an API. It is built in Rust, works with 15+ providers including Anthropic, OpenAI, Google, Ollama and OpenRouter (or your existing Claude, ChatGPT or Gemini subscriptions via ACP), connects to 70+ extensions through the Model Context Protocol, and is part of the Agentic AI Foundation at the Linux Foundation. Best for: Developers and knowledge workers who want one local agent for code, research, writing and automation with the model provider of their choice. Needs: macOS, Linux or Windows; An LLM provider API key, a local Ollama model, or an existing subscription via ACP. Start here: [Install the Goose CLI and hand it a real task](https://dedupedai.com/use-cases/aaif-goose-goose/install-goose-cli-and-run-a-task/) (Beginner, 20 minutes) 1. Install the CLI with curl -fsSL https://github.com/aaif-goose/goose/releases/download/stable/download_cli.sh | bash 2. Follow the Quickstart to configure a provider such as Anthropic, OpenAI or Ollama 3. Start a session in your project folder 4. Describe a task such as writing a script or migrating code All 3 use cases: https://dedupedai.com/use-cases/aaif-goose-goose/ | GitHub: https://github.com/aaif-goose/goose ## Questions ### What is the best open-source AI coding agent? The best open-source AI repos for AI coding agents right now are Claude Code, OpenCode and Gemini CLI. All three run in your terminal and are the most starred coding agents we track; Cline also works inside your IDE, and goose runs as a desktop app, a CLI and an API. ### Which repo for AI coding agents is easiest to start with? Gemini CLI: its starter use case "Ask Gemini to explain a codebase from the terminal" is rated Beginner and takes about 10 minutes. ### How do I get started with Claude Code? Install Claude Code with one of the README methods, for example curl -fsSL https://claude.ai/install.sh | bash on macOS or Linux; Navigate to your project directory; Run claude. Full steps and a copy-paste agent prompt: https://dedupedai.com/use-cases/anthropics-claude-code/explain-unfamiliar-codebase/ ### Are these AI coding agents repos free? They are open-source projects on GitHub (check each license; some, like fair-code projects, limit commercial use). Some also sell a hosted version or need a paid model API key; the requirements for each repo are listed on this page. ### How is this list of AI coding agents repos chosen? Deduped tracks the hottest AI repos on GitHub and writes step-by-step use cases from each README. For this page we picked the repos that do this job, ordered by fit and then GitHub stars. Repos we only inferred from a short README, or that we do not index, are left out. The page is rebuilt with the site; stars are as of the last build in 2026. Third-party repos we don't own or vouch for. AI-written use cases can be wrong: review everything and run at your own risk.