lockstep
4 min readNaman Jain

Claude Code alternatives: 7 coding agents compared

Claude Code alternatives compared from their docs: Codex CLI, Cursor, GitHub Copilot, Google Antigravity, Kiro, OpenCode, and Gemini CLI, and who each fits.

A developer likes Claude Code but wants to try something else: a different model, an open source tool, or whatever the team down the hall uses. The alternatives look similar from a distance. Up close, they differ in where they run, which models they use, and how they're configured.

This guide summarizes seven, from each tool's own documentation, checked on 16 November 2026, and links to the full comparisons. It leaves out prices and benchmarks.

At a glance

ToolModelsShapeReads AGENTS.md
Codex CLI (OpenAI)OpenAITerminal agent, plus ChatGPT desktop appYes
CursorSeveral providersAI code editor, cloud agents, CLIYes
GitHub CopilotSeveralIDE extensions, cloud agent, CLI, app, GitHub webYes, for agent features
Google AntigravityGemini 4, plus Claude on some plansDesktop app, agy CLI, IDE extensions, SDKYes
KiroSeveral providers, Auto routingIDE, CLI, web, mobile; spec-drivenYes
OpenCode200+ providersOpen source desktop, terminal, mobile, SDKYes, only
Gemini CLI (Google)GeminiTerminal agent, now for enterprise Code AssistUses GEMINI.md

Codex CLI

OpenAI's terminal agent. It reads AGENTS.md, supports skills in .agents/skills, MCP, hooks, and subagents. Choose it if you're in the ChatGPT ecosystem. See Codex memory and Claude Code vs Cursor vs Codex.

Cursor

An AI-first code editor with rules, skills, hooks, and subagents, Cloud Agents that run in isolated VMs, and Bugbot for pull request review. Choose it if you want the AI inside your editor and model choice. See Cursor vs GitHub Copilot.

GitHub Copilot

Built into GitHub: inline suggestions across VS Code, Visual Studio, JetBrains, Eclipse, and Xcode, a cloud agent you assign issues to, code review, a CLI, the Copilot app, and Copilot Memory. Choose it if your work already runs through GitHub issues and pull requests. See AGENTS.md vs copilot-instructions.md.

Google Antigravity

Google's agent platform: a desktop app for orchestrating agents, the agy CLI, extensions for six IDEs, and a Python SDK, centered on Gemini 4 with an Orchestrator option. Rules in AGENTS.md, GEMINI.md, or .agents/rules/. Choose it for Gemini models or Google Cloud. See Cursor vs Antigravity vs Claude Code.

Kiro

A coding agent built around specs: requirements, design, and tasks files for each feature, plus steering files for standing context. Available as an IDE, CLI, web, and mobile, with several model providers. Choose it if you want spec-driven development built in. See Kiro vs Claude Code.

OpenCode

Open source under the MIT License, working with free models or 200+ providers, from a desktop app, terminal, mobile app, or SDK. Reads AGENTS.md only, and loads skills from .claude/skills for compatibility. Choose it for open source or provider freedom. See OpenCode vs Claude Code.

Gemini CLI

Google moved free and Google AI Pro and Ultra users to Antigravity CLI on 18 June 2026. Gemini CLI remains for Gemini Code Assist Standard and Enterprise customers and paid API keys. See Gemini CLI vs Claude Code vs Codex CLI.

Choosing by what you want to change

You wantTry
Other model providersOpenCode, Cursor, Kiro
Open sourceOpenCode
Work tied to GitHub issues and PRsGitHub Copilot
OpenAI modelsCodex CLI
Gemini modelsAntigravity
Specs built inKiro
An AI-first editorCursor

Switching without starting over

The ecosystem has converged on shared formats:

  1. Put shared instructions in AGENTS.md. Most of these tools read it. Import it into CLAUDE.md with @AGENTS.md if you keep Claude Code.
  2. Keep skills in the Agent Skills format, in a folder the tools read or symlinked to one.
  3. Reuse MCP servers. Every tool here supports MCP; only the config file differs.

What doesn't transfer

Instruction files and skills move between tools. The decisions your team made last month don't live in either, unless someone wrote them down where every agent can read them.

Where Lockstep fits

Lockstep captures decisions from PRDs, Slack, and pull requests, routes each to its owner for approval, and briefs Claude Code sessions automatically. Briefings, hooks, and managed skills are Claude Code features, and Codex isn't supported; other MCP clients can call Lockstep's tools directly.

See how it works.

Sources checked 16 November 2026

For Claude Code's own options, see Claude Code desktop vs CLI vs VS Code. For getting more from Claude Code before switching, see Claude Code best practices. For general-purpose agents, see Claude Code vs OpenClaw vs Hermes Agent.

Frequently asked questions

What are the best Claude Code alternatives?
It depends on what you want to change. For other model providers, OpenCode or Cursor. For a GitHub-native workflow, GitHub Copilot. For OpenAI models, Codex CLI. For Gemini models, Google Antigravity. For built-in spec-driven development, Kiro.
Is there an open source alternative to Claude Code?
Yes. OpenCode is an MIT-licensed coding agent that works with free models or 200+ providers, from a desktop app, terminal, phone, or SDK. It reads AGENTS.md and also loads skills from .claude/skills.
Can I switch from Claude Code without rewriting my setup?
Mostly. Codex, Cursor, Antigravity, Copilot's agents, Kiro, and OpenCode read AGENTS.md, and most support Agent Skills. Keep shared instructions in AGENTS.md, import it into CLAUDE.md with @AGENTS.md, and keep skills in a folder the others read.
What happened to Gemini CLI?
Since 18 June 2026, Gemini CLI serves Gemini Code Assist Standard and Enterprise customers and paid API keys. Free and Google AI Pro and Ultra users moved to Antigravity CLI, according to Google.
Which Claude Code alternative has the best team controls?
Claude Code's managed settings are among the most detailed. GitHub Copilot offers organization policies, file exclusions, and audit data through GitHub. OpenCode has policies deliverable from its Console. Check each against the specific controls your security team needs.

Keep reading