Agentic Coding Tools for Claude Code & Codex

A fully open-source coding agent for the terminal, model-agnostic so you can point it at Claude, GPT, or local models. Positioned as a free alternative to proprietary agents with the same edit-run-iterate loop, it has become one of the most-starred agentic coding projects on GitHub.

Anthropic's official agentic coding tool. It lives in your terminal, reads your whole codebase, and edits files, runs commands, and opens pull requests from plain-English instructions. The reference implementation for what a Claude-powered coding agent can do, and the foundation most tools in this list extend.

Google's open-source AI agent that brings Gemini into your terminal. It handles coding, file edits, and shell tasks through an agent loop, and ships a large-context workflow aimed at reasoning over big repositories.

OpenAI's lightweight, open-source coding agent that runs in your terminal. Codex plans tasks, edits code, and executes commands locally, and is the OpenAI-side counterpart to Claude Code. A huge ecosystem of skills, routers, and orchestrators has grown around it.

A lightweight coding agent built to run open models like DeepSeek, Kimi, and Qwen. Open Interpreter lets a language model execute code on your machine through a natural-language chat, pioneering the local 'code interpreter in your terminal' pattern.

An autonomous coding agent available as an SDK, an IDE extension, and a CLI. Cline plans multi-step changes, asks permission before running commands, and works across editors — a popular open-source option for developers who want an agent embedded directly in VS Code.

An open, extensible AI agent that goes beyond suggestions — it installs dependencies, executes code, edits files, and tests its own work. Built around a pluggable extension system so you can wire it to any toolchain or model.

Alibaba's open-source coding agent that lives in your terminal, tuned for the Qwen model family. It offers the familiar agentic edit-and-run loop with strong multilingual and large-context support.

An open-source agent that gives you a whole dev team of AI personas inside your code editor. Roo Code (a popular Cline fork) supports custom modes, multiple models, and granular approval controls for autonomous edits.

An all-in-one agentic engineering platform combining the best features of Cline and Roo Code with its own marketplace. Kilo focuses on letting teams build, ship, and iterate faster with model-agnostic coding agents.

An open-source terminal agent designed for large projects and real-world tasks. Plandex plans changes across many files, keeps a version-controlled sandbox of pending edits, and lets you review diffs before applying them.

GitHub's official CLI that brings the Copilot coding agent into your terminal. It can read repositories, propose and apply changes, and run commands, extending Copilot beyond the editor into agentic workflows.

A pioneering CLI that turns a natural-language spec into a full codebase, asking clarifying questions as it goes. gpt-engineer was the precursor to Lovable and remains a canonical example of end-to-end agentic code generation.

An open-source 'agentic software engineer' — give Devika a high-level objective and it breaks it into steps, researches, writes code, and iterates. One of the first community answers to Devin.

From Princeton, an agent that takes a GitHub issue and autonomously fixes it using the model of your choice. SWE-agent introduced the agent-computer interface that lets an LLM navigate, edit, and test a real repository, and is a standard on the SWE-bench leaderboard.

LangChain's open-source asynchronous coding agent. Open SWE runs long-horizon tasks in the background, planning and executing changes against a repository without needing you to babysit each step.

A radically simplified, ~100-line agent that solves GitHub issues or helps from the command line. mini-SWE-agent strips the concept down to its essentials, making it the clearest teaching example of how an autonomous coding agent actually works.

A small agent that writes genuinely useful code by generating tests first, then looping on the implementation until the tests pass. Its test-driven feedback loop keeps the output grounded and verifiable.

A multi-harness agentic plugin marketplace spanning Claude Code, Codex CLI, Cursor, OpenCode, and GitHub Copilot. It packages dozens of specialized subagents and commands you can drop into your agent of choice.

A kanban board for orchestrating coding agents. Vibe Kanban lets you queue tasks for Claude Code, Codex, or any agent and watch them work in parallel, turning a backlog into autonomous execution.

A terminal built for AI coding agents, with vertical tabs and notifications so you can run and monitor several Claude Code or Codex sessions at once without losing track of what each one is doing.

An autonomous agent loop that re-runs a coding agent repeatedly until every item in a product spec is complete. Ralph popularized the 'just keep looping until done' pattern with intelligent exit detection.

A code editor designed for the agent era — run an army of Claude Code, Codex, and other agents on your machine from one interface, reviewing and steering their work as a fleet.

A terminal manager for running multiple AI agents — Claude Code, Codex, OpenCode, and Amp — side by side in isolated git workspaces, so several agents can work on different tasks simultaneously.

An orchestration layer that coordinates coding agents like Claude Code and Codex, distributing tasks across them and consolidating the results so you drive several agents as one.

An open-source agentic development environment (YC-backed) for running multiple coding agents in parallel, each in its own sandbox, with a clean UI for reviewing and merging their output.

A cross-platform desktop control center for managing Claude Code, Codex, OpenCode, Gemini and other agents from one place — switching providers, keys, and configurations without editing config files by hand.

A router that lets you use Claude Code as a foundation while choosing which model actually serves each request. It enables swapping in different or cheaper models behind the Claude Code interface.

A powerful GUI app and toolkit for Claude Code. Opcode lets you create custom agents, manage interactive sessions visually, and track usage — a graphical front end for people who'd rather not live in the terminal.

A mobile and web client for Codex and Claude Code with realtime voice, end-to-end encryption, and full feature parity, so you can drive your coding agents from your phone.

A web and mobile interface for Claude Code, OpenCode, Cursor CLI, and Codex. It exposes your local agents through a browser so you can review and continue sessions from any device.

A pre-indexed code knowledge graph that auto-syncs as your code changes, feeding accurate structural context to Claude Code, Codex, Gemini, and Cursor so agents reason over real relationships instead of guesses.

An MCP toolkit that gives coding agents semantic code retrieval and precise symbol-level editing via language servers. Serena lets agents like Claude Code understand and modify large codebases far more accurately than plain text search.

A code-search MCP server that makes your entire codebase available as context to any coding agent. It indexes the repo with vector search so the agent retrieves only the relevant files for each task.

Microsoft's framework for building agentic AI through conversations between multiple specialized agents. AutoGen is a foundational toolkit for orchestrating LLMs, tools, and humans into collaborative multi-agent systems.

A framework for orchestrating role-playing autonomous agents that collaborate as a 'crew' to tackle complex tasks. CrewAI is one of the most widely used Python libraries for building multi-agent applications.

Google's open-source, code-first Python Agent Development Kit for building, evaluating, and deploying sophisticated AI agents, with first-class support for tools, evaluation, and deployment.

An agent framework that brings Pydantic's type-safety and structured-output discipline to LLM applications, making it straightforward to build reliable, well-validated agents in Python.