Building this in Claude Code?
Read up on Claude Code and see the apps people shipped with it.
Why this matters
Claude Code is the reference point for terminal coding agents, so people compare everything to it. But there are good reasons to look elsewhere: you want a free tier (Claude Code is not on the free plan), you want to use models from more than one vendor, you want open source, or your usage keeps hitting plan limits.
The field changed fast in 2026. Google retired Gemini CLI for most users in favor of the Antigravity CLI on June 18, 2026. SpaceX closed a $60 billion acquisition of Cursor's maker on August 14, 2026, and Cursor's plans now lead with Grok. GitHub's Copilot moved to credit-based billing. OpenAI's Codex CLI and the open-source OpenCode kept shipping.
How the picks were made: price to start, free tier, who it fits, lock-in, and how fast it moves. Each pick below is judged as a replacement or companion for Claude Code, so the notes cover what carries over when you switch.
This is a snapshot from November 2026. Nobody can predict 2027, and prices and limits change, so each pick says when to reconsider. The order is editorial, not a benchmark. For the full field, including Claude Code itself, see Best AI coding agents for 2027.
The short list
| Pick | Best for | Starts at (as of Nov 2026) | Skip it if |
|---|---|---|---|
| OpenAI Codex CLI | Closest open-source match | Limited on Free and Go, Plus $20/month | You want non-OpenAI models |
| OpenCode | Any model, no lock-in | Free, with free models included | You want one vendor's polish |
| Cursor | Editor-first with agents | Free, Plus $8 (Grok only) | You want Claude models cheaply |
| Antigravity CLI | Trying an agent for free | Free tier (weekly quota) | You need open source |
| GitHub Copilot CLI | GitHub-centered teams | Free, Pro $10/month | You want predictable usage |
| Aider | Git-native and minimal | Free (bring your own API key) | You want the newest agent features |
| goose | General-purpose, open governance | Free (bring your own model) | You want a coding-only tool |
What carries over from Claude Code
Most of your setup is portable if you keep it plain:
- Rules. Claude Code reads
CLAUDE.mdand also supportsAGENTS.md. Codex and Cursor readAGENTS.md. Keep shared rules there. - MCP servers. Codex, OpenCode, Cursor CLI, Copilot CLI and goose all list MCP support.
- Skills. Codex, OpenCode, Copilot CLI and Antigravity CLI list skill support.
- Hooks. Claude Code's hooks are its edge. Copilot CLI and Antigravity CLI list hooks, but treat each tool's hook format as its own.
Pick 1: OpenAI Codex CLI
A local coding agent from OpenAI, Apache-2.0. Sign in with a ChatGPT account (Plus, Pro, Business, Edu, Enterprise) or use an API key. It has /permissions for sandbox control, /init to create an AGENTS.md, codex mcp, skills, plugins, subagents and codex cloud for cloud tasks.
Use it when you want the nearest equivalent to Claude Code with open-source code.
Skip it when you want Claude models in the same tool.
Pricing: Free and Go ($8) get limited Codex in the desktop app. Plus is $20/month with CLI, web and IDE access. Pro starts at $100. Usage limits are estimates that depend on model and task size.
Weakness: limits are ranges, so heavy use is hard to forecast.
Reconsider if OpenAI's model lineup or plan limits change.
Pick 2: OpenCode
MIT-licensed, with a terminal UI, desktop app and SDK. Model access comes from 200+ providers through Models.dev, and some free models need no account. You can sign in with ChatGPT Plus/Pro or GitHub Copilot. It has build and plan agents, a general subagent, MCP, skills in SKILL.md files, custom slash commands and /undo.
Use it when you want to mix models and avoid lock-in.
Skip it when you want a single vendor's tuned experience.
Pricing: free to use, or OpenCode Go from $10/month for curated models.
Weakness: results depend on the model you choose, and the project moves fast.
Reconsider if hosting, maintainers or pricing change.
Pick 3: Cursor
An AI editor with an Agent, checkpoints, a coordinator for bigger jobs, and a CLI that runs the same agents in your terminal and in GitHub Actions. Its model list includes Grok, Claude, GPT and Gemini options. See Cursor vs Claude Code.
Use it when you want to review changes in an editor.
Skip it when you want Claude models on the entry plan. Plus ($8, introductory) is Grok only, and third-party models bill at API list price from Premium ($30).
Pricing: Free, Plus $8, Premium $30, Super $100, Ultra $200, Teams from $30 per user.
Weakness: SpaceX closed the acquisition on August 14, 2026, and the plans now favor Grok.
Reconsider if model access or prices change.
Pick 4: Antigravity CLI
Google's agy terminal agent, which replaced Gemini CLI for Google AI Pro, Ultra and free individual users. It lists skills, hooks, subagents and extensions at launch, and Google says it will not have one-to-one feature parity with Gemini CLI right away. See the Google Antigravity guide.
Use it when you want a free way to try a terminal agent.
Skip it when you need open source. Gemini CLI remains Apache-2.0 for API key and enterprise users, but the Antigravity CLI repo appears to hold only docs.
Pricing: free quota that refreshes weekly, then Google AI Pro and Ultra for higher limits. Third-party models are Ultra only. See Google's plans page for prices.
Weakness: quotas can change, and some early commenters reported hitting weekly limits fast.
Reconsider after the next quota change.
Pick 5: GitHub Copilot CLI
GitHub's terminal agent. Plan mode with Shift+Tab, custom instructions, MCP, custom agents, hooks and skills, plus working with issues and pull requests from the terminal. It can also use your own provider through environment variables, including OpenAI-compatible endpoints, Azure OpenAI, Anthropic and local models.
Use it when your team lives on GitHub.
Skip it when you need flat usage. Billing is in AI credits, one credit equals $0.01.
Pricing: Free, Pro $10, Pro+ $39, Max $100 per user per month. CLI included on those plans.
Weakness: credits mean cost depends on model and context size.
Reconsider if credit rates or included amounts change.
Pick 6: Aider
Open source (Apache-2.0) and git-native: it commits each edit and has /undo. A repo map gives the model an overview of your project. Pull your rules in with --read CONVENTIONS.md. See Aider vs Claude Code.
Use it when you want small and transparent with your own API key.
Skip it when you want modern agent features. The latest PyPI release is 0.86.2 from February 2026.
Pricing: free software, you pay for your model provider.
Weakness: slower release pace, and docs examples lean on older models.
Reconsider when a new release ships.
Pick 7: goose
A general-purpose open-source agent (Apache-2.0) from the Linux Foundation's Agentic AI Foundation. It runs as a desktop app or CLI, works with 15+ providers, can use existing Claude, ChatGPT or Gemini subscriptions through ACP, and connects to 70+ extensions over MCP.
Use it when you want an agent for more than code, with open governance.
Skip it when you want a tool built only around coding workflows.
Pricing: free software, bring your own model or subscription.
Weakness: broad scope means less coding-specific polish.
Reconsider if the project's governance or provider support changes.
Common mistakes
Switching for one feature. Most tools copy each other within months. Switch for price, models or openness, not a single feature.
Moving your rules into a tool-specific format. Keep shared rules in AGENTS.md so you can change tools without rewriting them.
Comparing tools on different tasks. Run the same real task from your repo in two tools and compare tests passed, files touched and cost.
What's next
Read the Claude Code power user guide to see what you would be giving up (hooks, subagents, MCP), then pick one alternative and run one real task on it. For the wider field, see Best AI coding agents for 2027. If you are comparing models rather than tools, read Claude vs ChatGPT for coding.
Last verified: November 2026.