What is the difference between Claude Code and Cline?
Claude Code is Anthropic's coding agent and runs only on Claude models; Cline is an open source coding agent that works with almost any model provider you connect. Both read your code, edit files and run commands under approval settings you control, and both now run in the editor and the terminal. The main differences are model choice, pricing model and how each handles planning and approvals.
If you want one vendor, one subscription and the deepest tie-in with Claude, Claude Code fits. If you want to pick your own models, including local or open weight ones, and pay per use, Cline fits.
How does Claude Code work?
Anthropic's overview describes Claude Code as "an agentic coding tool that reads your codebase, edits files, runs commands, and integrates with your development tools. Available in your terminal, IDE, desktop app, and browser."
You install it, run claude in a project, and describe a task. It explores the code, proposes edits and asks before acting, depending on your permission mode. The same engine runs in the terminal, a VS Code extension (which also installs in Cursor), a JetBrains plugin, a desktop app and at claude.ai/code.
Key pieces:
- Models. Claude only. On Pro, Max, Team, Enterprise and the Anthropic API the default is Opus 5.5, per the model configuration docs. You can switch to Sonnet or Haiku with
/model. - Permissions. Six permission modes, from
default(shown as Manual, which "stops and asks you before most actions that edit files, run shell commands, or reach the network") toautoandbypassPermissions. On v2.1.283 or later, interactive terminal and VS Code sessions start inauto, where a classifier reviews actions instead of you. - Customization. CLAUDE.md instructions, skills, hooks, subagents, MCP servers and plugins.
- Automation.
claude -pfor scripts, plus GitHub Actions and GitLab CI/CD integrations.
How does Cline work?
Cline's docs describe it simply: "Cline is an AI coding agent that lives in your editor and your terminal." The GitHub repo calls it "The open source coding agent in your IDE, terminal, & desktop!" and it is licensed Apache-2.0.
You install the extension in VS Code, Cursor, Windsurf, VSCodium, Antigravity or a JetBrains IDE, then choose a provider. There is also a CLI (npm i -g cline), a desktop app for open weight models, an SDK and cloud agents that work on a GitHub repo in a sandbox.
Key pieces:
- Models. Bring your own. Cline documents providers including Anthropic, OpenAI, Google Gemini, OpenRouter, Amazon Bedrock, DeepSeek, any OpenAI-compatible API, and local models through Ollama or LM Studio, plus a reference list of 30+ other providers.
- Plan and Act. In Plan mode, "Cline can read your codebase, run searches, and discuss strategy, but cannot modify any files or execute commands." You switch to Act mode to carry out the plan with the same context.
- Approvals. The README says "Every file edit and terminal command requires your approval," unless you turn on Auto Approve, which you set per category: reading, editing, safe commands, all commands, browser and MCP.
- Checkpoints. "Every time Cline modifies a file or runs a command, it saves a snapshot of your project files," kept in a shadow Git repo so your own history stays clean.
Claude Code vs Cline side by side
| Claude Code | Cline | |
|---|---|---|
| Made by | Anthropic | Cline Bot Inc., open source (Apache-2.0) |
| Models | Anthropic's Claude models | Your choice of provider, including Claude, GPT, Gemini, open weight and local models |
| Where it runs | Terminal, VS Code and Cursor extension, JetBrains, desktop app, web | VS Code, Cursor, Windsurf, VSCodium, Antigravity and JetBrains extensions, CLI, desktop app, cloud agents |
| How you pay | Paid Claude plan, or a Claude Console account billed per token | Your own API key, Cline credits pay-as-you-go, or ClinePass |
| Planning | plan permission mode |
Plan and Act modes |
| Safety default | Starts in auto mode, where a classifier reviews actions; Manual mode asks before edits, commands and network | Asks before every file edit and terminal command |
| Undo | Checkpoints with /rewind |
Checkpoints in a shadow Git repo |
| Project instructions | CLAUDE.md, or AGENTS.md when there is no CLAUDE.md |
.clinerules/, .cline/rules/, AGENTS.md, .cursorrules, .windsurfrules |
Skills (SKILL.md) |
.claude/skills/, ~/.claude/skills/ |
.cline/skills/, .clinerules/skills/, .claude/skills/, ~/.cline/skills/ |
| Extension points | Hooks, subagents, MCP, plugins | Hooks, subagents, MCP, plugins |
| Scripting | claude -p |
cline "task", headless with --json or piped input |
How do pricing and model access compare?
Claude Code is part of every paid Claude plan. As of October 2026, claude.com/pricing lists Pro at $20 a month ($17 a month billed annually), Max from $100 a month, and Team standard seats at $25 per seat monthly or $20 billed annually. The free plan does not include Claude Code. A Claude Console account bills per token instead, and teams can run it through cloud providers such as Amazon Bedrock.
Cline is open source, so the cost is model usage: you pay whoever serves the model. Cline offers three routes:
- Bring your own key from Anthropic, OpenAI, Google or another provider, and pay that provider's API rates.
- Cline (usage-billing), where you "sign in once, add Cline credits, and select from available models directly in Cline." Some models are tagged FREE in the selector.
- ClinePass, $9.99 a month as of October 2026, for open coding models such as GLM, Kimi, DeepSeek and Qwen.
The trade-off is plain. A Claude subscription gives a predictable bill with usage limits. Cline lets you pick a cheaper or local model for routine work and a stronger one for hard tasks, but per-token bills can vary.
How do rules, skills and MCP compare?
Rules. Claude Code reads CLAUDE.md at the start of every session, and its memory docs say it can read AGENTS.md in place of CLAUDE.md. Cline's rules page lists .clinerules/ and .cline/rules/ folders, and also detects .cursorrules, .windsurfrules and AGENTS.md. Each rule gets a toggle in Cline's Rules panel. If you use both tools, put shared rules in AGENTS.md and import it from CLAUDE.md with an @AGENTS.md line.
Skills. Both support Agent Skills, folders with a SKILL.md file that load on demand. Cline's skills docs require name to match the folder and cap description at 1,024 characters, the same rules as the open standard. Cline recommends .cline/skills/, but its docs also list .claude/skills/ as a project location, so a skill committed there works in both tools. Global folders differ: ~/.claude/skills/ for Claude Code and ~/.cline/skills/ for Cline. See the Claude Code skills guide for more.
MCP. Both connect to Model Context Protocol servers for outside tools and data. See the Claude Code MCP guide for its setup.
Which should you pick?
Pick Claude Code if:
- You already pay for Claude Pro, Max or Team and want coding included.
- You want Anthropic's models and features built for them, such as skills, hooks and permission modes, with no provider setup.
- You want the same agent in your terminal, IDE, desktop app and browser.
- You prefer a flat subscription with usage limits over per-token billing.
Pick Cline if:
- You want to choose your model, switch providers, or run local and open weight models.
- You want open source code you can read and audit.
- You like an explicit Plan then Act flow with approval on each edit and command.
- Your team uses several editors, including Windsurf or VSCodium.
Neither is a wrong choice. Both are actively developed agents with similar building blocks, so run one real task from your backlog through each before you commit.
Keep your prompts portable across agents
The prompts and skills you refine are worth more than any one agent. Kipt keeps them as Markdown in a private GitHub repo you own, and installs every skill into Claude Code or Codex with one npx skills add command. To turn a prompt you reuse into a skill folder, try the free SKILL.md generator.