Cline vs Claude Code: pricing, models, control

Cline lives in your editor; Claude Code started in your terminal and now lives everywhere. Both edit files, run commands and ask before doing anything destructive. Here is how they differ on the things that decide a team's choice, taken from the two companies' own pages as of September 2026.

At a glance

• Cline: Apache 2.0 open source from Cline Bot Inc., about 67k GitHub stars. VS Code extension, JetBrains plugin, and a CLI with interactive and headless modes for CI. Free for individual developers; you pay only for inference, with your own keys for Anthropic, OpenAI, Google, OpenRouter, Bedrock, Vertex, Groq, Cerebras, DeepSeek and others. Enterprise is custom, with SSO, an SLA and dedicated support.
• Claude Code: Anthropic's proprietary agent. Terminal, VS Code, JetBrains, desktop app, web, mobile hand-off. Billed through a Claude subscription (Pro $20 a month or $17 annually; Max from $100 with 5x or 20x usage; Team seats $20 annually or $25 monthly, premium seat $100) or API pay-as-you-go through the Console.

How you pay

Cline's model is usage-based with no subscription: "You only pay for AI inference credits when you use AI models." That means your cost is the provider's token price, and a heavy week is a heavy bill. Claude Code on a subscription is the opposite shape: a flat fee with a usage limit, so the failure mode is hitting the limit rather than a surprise invoice, unless you are on API billing, where it behaves like Cline.

Model choice

Cline is model agnostic; a single day of tasks can span several providers and a local model. Claude Code runs Anthropic models, on Anthropic's surfaces or through third-party hosting from the terminal and IDE extensions. If you want to route by task, Cline gives you the dial; if you want the best integration with one model family, Claude Code is built around it.

Approval and control

• Cline: Plan mode explores and strategises, Act mode edits, and in Act mode "every file edit and terminal command requires your approval" until you flip auto-approve. Checkpoints let you undo the agent's work as diffs.
• Claude Code: permission prompts per tool, plus hooks that run your own shell command before or after an action and can block it, which is how external gates plug in. Subagents, background agents and routines extend it from one session to many.

Which to pick

• Pick Cline if you want an open source agent inside VS Code or JetBrains, to choose your provider per task, and to pay the provider directly.
• Pick Claude Code if you already pay for Claude, want the terminal-first workflow with subagents and hooks, or want the desktop, web and mobile surfaces.

What either one actually cost you

Both keep a local record of every task. ClawMetry reads Cline's task history per project (instruction, replies, every tool call with arguments and result) and computes cost per task across whichever providers the task touched, cache-aware. For Claude Code it does the same per session, with the sub-agent tree and per-turn cache-read accounting. Nothing to instrument; both appear next to each other on the runtimes page. Details on the Cline page and the Claude Code page.

Sources

• Cline pricing, cline/cline on GitHub
• Claude Code overview, Claude pricing