LaunchedEditorial Listing

Cline

Cline · Open-Source AI Coding Agent for VS Code, JetBrains & Terminal

Open Cline

Cline is a free, Apache 2.0 AI coding agent that plans changes, edits files across your project, and runs terminal commands, asking for your approval at each step by default. It runs as a VS Code extension (and in forks like Cursor and Windsurf), a JetBrains plugin, a CLI, and a beta desktop app, and suits developers who want to choose their own model. The tradeoff is cost and control: the software is free, but you pay per token for the model you connect, so cost depends on the model and how long tasks run.

PricingFreemium
Setupeasy
Runs onDesktop
APIYes
Open sourceYes
DocsYes
CategoryCoding
Open SourceCodingVS CodeJetBrainsAgentic CodingCLIMCP

Best for

Developers and teams who want an open-source, model-agnostic AI coding agent inside their existing IDE, with step-by-step approval and per-task cost visibility

Not ideal for

Non-technical users, or anyone who wants a flat-price assistant with a bundled model and no per-token cost to watch

Who it's for

Developers who want an open-source agentic coding assistant in their IDE or terminal, using their own model keys

Capabilities

  • Plan mode (reads code and proposes an approach, cannot edit files or run commands) and Act mode (executes), with an option to use a different model for each
  • Per-action approval by default, with per-category Auto Approve settings and a YOLO mode
  • Checkpoints: workspace snapshots in a shadow git repository after each tool use, with restore for files, task, or both
  • Edits code across multiple files, runs terminal commands with live output, and automates a browser
  • Bring your own model: Anthropic, OpenAI, Gemini, OpenRouter, Bedrock, Vertex, DeepSeek, 30+ other providers, any OpenAI-compatible API, or local models via Ollama and LM Studio
  • Per-task token and API cost tracking in the task header, including prompt-cache savings
  • Auto Compact, /smol, and /newtask commands to manage context length on long tasks
  • MCP server support and project-specific rules via .clinerules files
  • Runs in VS Code and forks, JetBrains IDEs (Early Access), a CLI for headless and CI use, an SDK, and Cline Desktop (beta) for macOS, Windows, and Linux
  • ACP support for Zed, Neovim, and Emacs, plus Cline Kanban (research preview) for parallel agents in git worktrees

Limitations

  • No bundled model: you pay your provider, Cline at provider cost, or ClinePass, unless you run local models
  • Long autonomous tasks can use a lot of tokens; cost depends on the model and how much context each task carries
  • The CLI auto-approves actions by default (except in ACP mode) and has no allow or deny list for shell commands
  • Local models need capable hardware; Cline's docs suggest 32GB+ RAM and recommend the compact prompt setting
  • Claude Pro/Max through the Claude CLI may not stream token by token, and image uploads and prompt caching are limited
  • Checkpoints can use significant storage in very large repositories
  • Telemetry is on by default (it excludes code and prompts except with the Cline and ClinePass providers) and must be turned off manually
  • Cline Desktop is in beta and Cline Kanban is a research preview

Use cases

  • Implementing a feature or refactor across multiple files from a single instruction
  • Running and fixing failing tests or build errors inside the editor
  • Automating repository tasks headlessly in CI/CD via the CLI
  • Extending the agent with MCP servers for internal tools, docs, or APIs
  • Using local models via Ollama or LM Studio to avoid per-request API costs

Cline pricing

Open Source (BYOK)

Free

  • Free and open source under the Apache 2.0 license, with no subscriptions or seat fees
  • Bring your own LLM API key and pay only your provider's usage
  • VS Code extension (and forks), CLI, and Desktop
  • Free JetBrains plugin (Early Access) on JetBrains Marketplace; also listed as an Enterprise plan feature

Cline provider

Pay-as-you-go (at provider cost)

  • Sign in instead of managing provider API keys
  • Billed at the model provider's price with no markup
  • Occasional limited-time free model promotions (usage may be used to improve the model)

ClinePass

$9.99/month

Billed monthly

  • 2–5x usage on curated open coding models compared with standard API rates
  • Usage limits over a 5-hour rolling window, weekly, and monthly
  • Optional; other providers remain available alongside it

Enterprise

Custom

  • Team collaboration and centralized billing
  • RBAC, SSO/OIDC, and authentication logs
  • VPC, on-prem, or air-gapped deployment and SLA
  • JetBrains extension

Free tier limits: The open-source extension is free for individuals; you pay only for model usage (or $0 per request with local models).

Note: Separate the software cost from the model cost. Cline itself is free (Apache 2.0) with no seat fees. Model usage is billed per token by your provider, by Cline at provider cost, or through a $9.99/month ClinePass for open coding models, and Cline displays each task's running cost. To keep model costs down, Cline's docs point to using a cheaper model for Act than for Plan, prompt caching on supported providers, compacting context or starting a new task when context grows, and local models. Enterprise pricing is custom via sales.

Technical specs

Modalities

Text, Code, Image

Available models

Anthropic Claude (API key, or Claude Pro/Max via the Claude CLI)OpenAI GPT models (API key or ChatGPT/Codex subscription sign-in)Google GeminiOpenRouter, AWS Bedrock, GCP Vertex AIDeepSeek, Qwen, MiniMax, xAI, Mistral, Groq, Cerebras, and 30+ other providersOllama and LM Studio local modelsAny OpenAI-compatible endpoint

Cline vs. competitors

Cline vs. OpenCode

Both are open-source and bring-your-own-model. Cline lives mainly in an editor side panel (VS Code and forks, JetBrains) with per-step approvals by default; OpenCode is terminal-first, with a desktop app and an editor split terminal. Cline can use a Claude Pro/Max subscription through the Claude CLI; OpenCode cannot.

Cline vs. Kilo Code

Kilo Code is also an open-source agent for VS Code, JetBrains, and the terminal with bring-your-own-key model access, and adds cloud agents and a Slack bot. Cline's distinguishing pieces are its Plan/Act split, checkpoints, and the at-cost Cline provider and ClinePass.

Cline vs. Cursor

Cursor is a VS Code fork with inline autocomplete and a subscription that bundles model usage. Cline is an extension you add to your existing editor, with no autocomplete and per-token model costs you control. Cursor is lower-friction; Cline gives model choice and an open codebase.

Cline vs. GitHub Copilot

GitHub Copilot bundles completions, chat, and an agent mode into a subscription with GitHub integration. Cline focuses on agentic, multi-step tasks with the model of your choice and per-task cost tracking, but has no inline autocomplete.

Frequently asked questions

How do I set up Cline?

Install the Cline extension from the VS Code Marketplace (Open VSX for Windsurf and VSCodium) or the plugin from the JetBrains Marketplace, then open Cline and pick an API provider in its settings. The options are the Cline provider (sign in with Google, GitHub, or email, no API key), ClinePass, a direct provider such as Anthropic, OpenAI, or OpenRouter (paste your API key), or a local model in Ollama or LM Studio (no key). Then choose a model. For the CLI, install Node.js 22 or later, run npm install -g cline, then cline auth.

Is Cline free, and how much do models cost?

The Cline software is free and open source, with no subscription or seat fee. You pay for model usage: your provider's per-token rates with your own key, the same rates through the Cline provider (no markup), or ClinePass at $9.99/month for open coding models. Cline shows token use and API cost for each task in the task header. Local models via Ollama or LM Studio have no per-request cost but need capable hardware.

How can I reduce Cline's model costs?

Cline's docs suggest several levers: set a different, cheaper model for Act mode than for Plan mode; use providers that support prompt caching (Cline tracks cache savings); keep tasks focused and use Auto Compact, /smol, or /newtask when context grows; consider ClinePass for open coding models; or run a local model. Watching the per-task cost in the task header shows which tasks are expensive.

Can I use my Claude Pro/Max or ChatGPT subscription with Cline?

Yes, with caveats. Cline can use a Claude Pro/Max subscription through the Claude CLI, but responses may not stream token by token, and image uploads and prompt caching are limited in that mode. For OpenAI, you can sign in with your OpenAI account through the Codex subscription option; the models available depend on your OpenAI plan.

What is the difference between Plan and Act mode?

In Plan mode, Cline can read your codebase, run searches, and discuss strategy, but cannot modify files or execute commands. In Act mode it carries out the plan. You can assign a different model to each mode, for example a stronger model to plan and a cheaper one to act.

Is it safe to let Cline run commands, and does it send my code to Cline?

In the editor, every file edit and command needs your approval unless you enable Auto Approve or YOLO mode, and checkpoints let you restore earlier workspace states. The CLI auto-approves by default (except in ACP mode), so review its settings before using it on important repositories. With your own API key, prompts go directly to your chosen provider; Cline's telemetry is on by default but excludes code and prompts except when you use the Cline or ClinePass providers, and you can opt out.

Integrations & fit

VS CodeCursorWindsurfVSCodiumJetBrains IDEsZed (ACP)Neovim (ACP)AnthropicOpenAIGoogle GeminiOpenRouterAWS BedrockOllamaLM StudioMCP serversGit
Good fit forSolo / individual, Startup / small team, Enterprise
Pricing modelFreemium· Free tier available
See pricing on Cline →

Alternatives to consider

See the best AI coding agents →

About Cline

Cline (Apache 2.0) installs from the VS Code Marketplace, from Open VSX for forks such as Windsurf and VSCodium (Cursor and Antigravity are also covered in the install guide), or from the JetBrains Marketplace (Early Access). It also ships as a CLI for headless and CI use on macOS, Linux, and Windows (npm install -g cline), an SDK, Cline Desktop (beta) for macOS, Windows, and Linux, ACP support for Zed, JetBrains IDEs, Neovim, and Emacs, and Cline Kanban, a research preview for running parallel agents in git worktrees. Work is split into Plan mode, where Cline reads the codebase and discusses an approach but cannot modify files or run commands, and Act mode, where it executes. By default every action needs your approval; Auto Approve settings and YOLO mode remove those checks per category or entirely. Checkpoints snapshot your workspace in a shadow git repository after each tool use so you can restore files, the task, or both. Cline is bring-your-own-model: on first run you pick an API provider in settings, either the Cline provider (sign in, no API key, billed at the model provider's price with no markup), ClinePass, a direct provider key (Anthropic, OpenAI, Gemini, OpenRouter, Bedrock, Vertex, DeepSeek, and 30+ others), or a local model through Ollama or LM Studio. You can also sign in with an OpenAI ChatGPT/Codex subscription, or use a Claude Pro/Max subscription through the Claude CLI, with limits on streaming, image uploads, and prompt caching in that mode. The software is free for individuals with no seat fees. Model usage is the cost: Cline shows token use and API cost for each task in the task header, ClinePass ($9.99/month) gives 2–5x usage on curated open coding models, and Enterprise (custom pricing) adds team management, SSO, and private deployment options.

Are you the founder? Claim this listing →