Claude Code Alternatives in 2026 (cto CLI Walkthrough)

If you're hunting for a Claude Code alternative, the question is usually one of three: you've hit Claude Code's usage limits, you want to run more than one model family from the same CLI, or you don't want to pay $20–$200/month. This page is a walkthrough of cto CLI - what it actually does, the features that matter when you're moving off Claude Code, and how it compares to the rest of the category.

TL;DR. cto CLI is an opencode plugin + provider - the agent loop, file editing, plan mode, worktrees, hooks, skills, slash commands, and MCP support are all open-source opencode running in your terminal. cto adds the cto.new gateway, which means you don't bring API keys - pick from Claude Opus 4.8, GPT-5.5, Gemini 3.1 Pro, GPT-5.3 Codex, GLM 5.1, Grok 4.20 and more via opencode's /model slash command. Free forever (ad-supported) with rolling 24-hour and 7-day usage limits. MCP support is opencode-native: bring your own MCP server configs.

What cto CLI is

cto is a single binary that drops into any terminal:

curl -fsSL https://cto.new/install | bash

Run cto from any project. First launch opens your browser to log you in; subsequent runs are silent. cto CLI ships as an opencode plugin + provider - the agent runs locally in your terminal, calls route through the cto.new gateway for LLM and MCP traffic. You inherit everything opencode does (file editing, plan mode, worktrees, hooks, skills, slash commands, MCP) and you don't bring keys.

The full command surface:

cto                        Launch the cto coding agent (opencode TUI)
cto login                  Re-authenticate this device
cto logout                 Revoke credentials, forget the local copy
cto upgrade                Update cto to the latest release
cto uninstall              Remove cto + ~/.cto from this machine
cto --version              Print the cto version
cto --licenses             Print license + third-party notices
cto --help                 Show usage

Sub-commands not listed above pass through to opencode. Inside the TUI, /cto-login and /cto-logout let you re-auth without quitting.

The features Claude Code refugees care about

Every frontier model, no API keys

Claude Code is Claude. cto CLI gives you the full frontier menu through opencode's /model slash command: Opus 4.8, Sonnet 4.6, GPT-5.5, GPT-5.3 Codex, Gemini 3.5 Flash, Gemini 3.1 Pro, GLM 5.1, Grok 4.20, Kimi K2.6, MiniMax M2.7. The cto.new gateway handles provider auth - you don't bring keys, you don't watch dashboards, you don't pay per token on the free tier.

Free forever (ad-supported), rolling 24h + 7d limits

The pricing model is genuinely free, not a trial. Limits are rolling 24-hour and 7-day windows - not a calendar-day reset. You see short inline ads in the CLI session as part of the free tier. Premium raises the limits and unlocks higher-performance models and premium agent features. No credit card, no API key, no per-token bill.

MCP that's already wired

MCP support is opencode-native. Configure your MCP servers (Linear, Sentry, Notion, custom - any stdio or streamable HTTP server) using opencode's standard MCP config and they'll be available to the agent. The cto wrapper doesn't replace or extend that - what opencode supports, cto CLI supports.

Built as an opencode plugin + provider

cto CLI distributes as two opencode components: a plugin that adds cto-specific behavior (auth, install/upgrade, ads layer) and a provider (the cto.new gateway) that handles LLM calls. Everything else is opencode: the TUI agent, file editing, plan mode, worktrees, hooks, skills, slash commands, LSP, multi-session support, native MCP. opencode is open source (Apache 2.0); cto is what makes it free-to-use without bringing keys.

How cto CLI compares to other Claude Code alternatives

Migrating from Claude Code

There's nothing dramatic - both tools work on any project directory.

# Install cto alongside Claude Code (they coexist fine)
curl -fsSL https://cto.new/install | bash

# In your project root
cto
# Browser opens to log you in - same Google/GitHub identity if you want.
# Drop into the TUI and start coding.

Project-rules files (e.g., your CLAUDE.md) are read by cto too if you point the agent at them - opencode supports the same convention. MCP servers you've configured in Claude Code can be wired into cto's account-wide MCP settings (each MCP server has the same JSON shape; you re-add them in the cto.new web UI rather than per-project files).

Why people pick cto CLI

1. One tool, every model, no keys. Pick from 10+ frontier LLMs via opencode's /model slash command. Claude Code's third-party provider story exists but the experience is still Claude-first, and you bring keys for the others.
2. Predictable pricing. Rolling 24h + 7d windows on the free tier; flat-rate premium with a higher window. No AI Credits, no per-token meter, no surprise bill.
3. opencode-native MCP. Whatever MCP server you want, configure it via opencode's standard format and the cto CLI picks it up. No custom config language; no per-CLI wrappers.

FAQ

Is Claude Code free?

The Claude Code binary installs free, but actually using it requires a paid Claude plan, an Anthropic Console balance, or a third-party provider you're paying. There's no genuinely-free path inside Claude Code itself.

Is cto CLI really free?

Yes - free forever, ad-supported. Free-tier limits are rolling 24-hour and 7-day windows (not a fixed midnight reset). Premium raises the limits and unlocks premium models + agent features.

Does cto CLI work with Claude?

Yes - Claude Opus 4.8 and Sonnet 4.6 are first-class. So are GPT-5.5, GPT-5.3 Codex, Gemini 3.5 Flash, Gemini 3.1 Pro, GLM 5.1, Grok 4.20, Kimi K2.6, and MiniMax M2.7. Pick whichever model you want from opencode's /model menu - the cto.new gateway handles provider auth.

How do I switch models in cto CLI?

Use opencode's /model slash command inside the TUI. The cto.new gateway exposes every frontier model on the menu; pick the one you want per session.

Does cto CLI support MCP?

Yes - natively via opencode. Configure your MCP servers in opencode's standard MCP config format (stdio or streamable HTTP) and the agent picks them up. The cto wrapper doesn't add or remove MCP capability; whatever opencode supports, cto CLI supports.

What's the relationship with opencode?

cto CLI distributes as an opencode plugin + provider. opencode is the open-source TUI agent (Apache 2.0) that handles file editing, command execution, the agent loop, plan mode, worktrees, hooks, skills, slash commands, LSP, and MCP. cto's plugin adds the cto-specific install/auth/upgrade layer; cto's provider is the cto.new gateway that handles LLM routing (so you don't bring API keys). Sub-commands not handled by cto pass through to opencode.

Next steps

• How MCP works in cto CLI →
• Best AI for coding 2026 →
• Free Claude Code: install cto in 30 seconds →
• Cursor CLI alternatives →