GitHub Copilot CLI got serious in 2025–2026 - it's multi-foundation-model now (Anthropic, Google, OpenAI), supports MCP shared with the VS Code extension, and has a free tier on Copilot Free. The wrinkle: GitHub moved to usage-based AI Credits billing in June 2026, which makes monthly cost a moving target depending on which model gets picked for each task and how many tokens it burns. This page covers cto CLI - same multi-model and MCP coverage with a different pricing shape (rolling 24h + 7d windows, no AI Credits math) - and the other CLIs in the space.
TL;DR. cto CLI is the "free forever with predictable limits" alternative to Copilot CLI's usage-based AI Credits model. Pick from 10+ frontier LLMs via opencode's /model slash command (Opus 4.8, GPT-5.5, Gemini 3.1 Pro, GPT-5.3 Codex, GLM 5.1, Grok 4.20, more) - the cto.new gateway handles provider auth so you don't bring keys. opencode-native MCP. No GitHub account dependency.
The pricing difference, concretely
Copilot CLI's AI Credits model means each task costs a different amount depending on:
- Which foundation model gets picked (Anthropic / Google / OpenAI)
- How many tokens it consumes (input + output + cached)
- Whether you're on Free, Pro ($10/mo), Pro+ ($39/mo), Business ($19/seat), or Enterprise ($39/seat)
- Whether your Credit allotment has been used this month
cto CLI's model is simpler:
- Free forever - ad-supported, rolling 24-hour and 7-day usage windows. Same usage costs the same amount because there's no per-token meter.
- Premium - flat-rate. Raised window + access to premium models + premium agent features.
If you're a heavy Copilot user past your AI Credit allotment, your monthly bill became unpredictable in June 2026. cto CLI's free tier is enough for most days; premium is the budget line if you're past it.
What cto CLI is
curl -fsSL https://cto.new/install | bash
cto ships as an opencode plugin + provider. The open-source TUI agent runs locally in your terminal - same agent loop, file editing, plan mode, worktrees, hooks, skills, slash commands, LSP, and native MCP. cto's plugin adds install/auth/upgrade and the ads layer; cto's provider is the cto.new gateway that handles LLM calls so you don't bring keys. Pick a model from the frontier menu via opencode's /model slash command.
Commands: cto, cto login, cto logout, cto upgrade, cto uninstall, cto --version, cto --licenses, cto --help. Sub-commands the wrapper doesn't recognize pass through to opencode.
Model switching - both let you flip per task
Copilot CLI lets you switch between Anthropic, Google, and OpenAI per task. cto CLI is the same: pick from any frontier model in the cto.new gateway via opencode's /model slash command. Switching is a keystroke in both tools - the difference is what you pay for it.
Models in the gateway: Claude Opus 4.8, Claude 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.
MCP - both support it
Copilot CLI shares MCP config with the VS Code extension (since March 2026). Useful if you're already on GitHub's MCP catalog.
cto CLI uses opencode's native MCP support. Configure MCP servers in opencode's standard MCP config format (stdio or streamable HTTP) and the agent picks them up - same as you would in vanilla opencode.
When Copilot CLI is still the right call
Honest cases where you stay on Copilot:
- You're deep in GitHub Actions / PR workflows. Copilot's native GH integration is unmatched. cto CLI runs in CI but doesn't have the same auto-PR-context integration.
- Your team is fully standardized on GitHub Enterprise. Identity flows, audit logs, etc. are cleanest on Copilot.
- Your usage is light enough that Copilot Free covers you. The free tier's caps are tight but real.
For everyone else - especially heavy users hitting the AI Credit math - cto CLI is the predictable-pricing alternative.
How to try side-by-side
# Install cto alongside Copilot CLI
curl -fsSL https://cto.new/install | bash
# Same project, run the same task in each
copilot ai "find and fix the failing tests"
cto # then in the TUI: find and fix the failing tests
# Pick your preferred model via /model - Sonnet 4.6, Opus 4.8, GPT-5.5, etc.
# Copilot picks based on your account's selected model.
Other CLIs in the space
- Claude Code - best Anthropic-native UX. Paid only.
- Cursor CLI - best Cursor editor parity. Tied to Cursor's Pro plan.
- Aider - open-source, BYOK, terminal-only.
- Codex CLI - OpenAI's open-source terminal agent (Apache 2.0).
- Gemini CLI - 1,000 free req/day on a Google account.
FAQ
Is GitHub Copilot CLI free?
Yes, on the Copilot Free plan, with tight usage caps. Pro ($10/mo), Pro+ ($39/mo), Business ($19/seat), and Enterprise ($39/seat) include larger AI Credit allotments. Overage is pay-as-you-go on listed API rates per model.
What changed with Copilot pricing in June 2026?
GitHub moved from request-based billing to usage-based AI Credits. Each plan includes a fixed Credit allotment; overage is pay-as-you-go per model. Same daily usage can cost different amounts depending on which model handled it.
Does cto CLI work in CI?
Yes - cto -p "prompt" works in scripts and CI runners. CI usage counts against your rolling 24h+7d window like any other usage; heavy CI throughput typically wants premium.
Does cto CLI support GitHub Actions?
You can run cto CLI inside a GitHub Actions workflow - the binary installs from curl -fsSL https://cto.new/install | bash. There's no first-party action yet; that's on the roadmap. For deep GH integration today, Copilot CLI's native action is stronger.
Which has more models?
cto CLI's gateway covers Claude (Opus 4.8, Sonnet 4.6), OpenAI (GPT-5.5, GPT-5.3 Codex), Google (Gemini 3.5 Flash, 3.1 Pro), plus GLM 5.1, Grok 4.20, Kimi K2.6, MiniMax M2.7. Copilot CLI covers Anthropic, Google, OpenAI - the three frontier providers, no broader open-model coverage.
Is cto CLI as well-integrated with GitHub as Copilot CLI?
No. Copilot's GitHub-native integration (PR context, repo auth, branch awareness via gh auth) is the deepest in the category. cto CLI works on any project but doesn't have the same auto-context-from-PR features. If GH integration is the reason you use Copilot, keep using Copilot.