Agent Rankの内訳
- 自律性
- 8
- 機能
- 6
- 連携
- 4
- 料金
- 10
- 完成度
- 8
- 検証性
- 10
自律性、機能、連携、料金、成熟度、編集部の検証を基に自動算出。デプロイごとに更新されます。 どのように算出されているか?
機能
- コード実行
- ツール利用
- ブラウザ操作
連携
- IDE
- VS Code
- Version control
- GitHub
料金プラン
- +MIT ライセンスで拡張機能を完全無料提供
- +API キーの持ち込み（BYO）
- +すべてのツールと承認モード
- +セルフホストに対応
Our take on Cline
Cline is the open-source coding agent that punches above its weight. Lives inside VS Code, brings your own LLM key, and gives you a level of transparency proprietary tools can't match.
- +Fully open source — audit every prompt, every tool call
- +Bring-your-own-model — works with Claude, GPT-4, Gemini, local models via Ollama
- +Inline diff UX in VS Code is the cleanest in the category
- +No per-task pricing — pay only your LLM provider
- −Setup is more involved than a one-click SaaS
- −No background / async mode — agent runs while VS Code is open
- −Quality is bounded by the model you choose; cheap models give cheap results
- −Documentation is improving but still lags polished commercial tools
- ·Engineers who want full control over the prompt and tool use
- ·Teams with corporate AI policies that disallow third-party SaaS
- ·Anyone already paying for Claude or GPT-4 API access
- ·You're responsible for the cost ceiling — runaway loops can burn $50 fast
- ·Self-hosted models work but give noticeably weaker results below 30B params
Architecture
Cline runs as a VS Code extension. It talks to your chosen LLM provider directly (no intermediate proxy), reads and writes files in your workspace, runs commands in the integrated terminal, and uses Playwright to drive a headless browser when needed.
Because there's no SaaS layer in between, every prompt, every tool result, and every edit is visible to you. This is a meaningful difference: when Cline misbehaves you can read the transcript and fix the prompt. With closed-source agents, you can't.
What you get for the open source price
Cline includes every major agent capability we look for:
- File edits with diff approval before each change
- Shell command execution with a confirm step you can disable per-command
- Browser automation for tasks like reading docs or filing GitHub issues
- Plan / Act mode toggle — sketch a multi-step plan first, then execute it
The only thing missing versus paid alternatives like Devin is background execution. Cline runs in your editor while you watch. Close VS Code, the agent stops.
Cost math
This is where Cline gets interesting. Using Claude Sonnet 4 for a typical 30-minute session — adding a feature, debugging, running tests — usually lands at $1-$3 of API spend. A heavy refactor across 20 files might hit $8. For most engineers, monthly cost sits between $20 and $80 — and you can hard-cap it via your provider's billing controls.
Compare to Cursor's $20/mo (capped requests) or Devin's $500/mo entry tier, and Cline ends up cheaper for moderate use, comparable for heavy use, and explicitly capped on the high end.
Where Cline really wins
Two scenarios where Cline is the right answer:
- You want to understand what the agent is doing. Cline's transparency is unmatched. You can debug the prompt itself.
- Corporate policy disallows agent SaaS. Self-hosted, your-keys-only, deployed via an enterprise VS Code marketplace — Cline ticks every compliance box that paid competitors can't.
Bottom line
Cline is the answer when you want a real agent without a vendor. Pair it with Claude Sonnet 4.7 and you get 85% of Devin's quality at 10-15% of the cost — but you give up background execution and pay in setup friction.
IDE 内で動作するオープンソースの自律的なコーディングエージェントです。 オープンソースプランから始めましょう。
よくある質問
Cline は本当に無料ですか？+
Cline 拡張機能自体は Apache 2.0 ライセンスの下で無料です — ソースの監査、フォーク、エアギャップ環境での実行が可能です。費用が発生するのは、持ち込む LLM の API キー（Anthropic、OpenAI、Gemini、または Ollama 経由のローカルモデル）です。クラウドモデルを日常的に使用する場合、トークン消費は月額 $40〜$120 程度です。既存のハードウェアでセルフホストする場合は $0 です。
Cline と Cursor の違いは何ですか？+
Cline は API キーを持ち込んで使用するオープンソースの VS Code 拡張機能です。Cursor は月額 $20 の定額制で提供される完成度の高いクローズドエディタです。Cline は透明性、モデルの柔軟性、コンプライアンス面で優れています。Cursor はポリッシュされた UX、統合されたタブ補完、予測可能なコストで優れています。
Cline はローカル LLM で動作しますか？+
はい — Cline は Ollama、LM Studio、vLLM、および OpenAI 互換のローカルエンドポイントと連携します。ローカルで実行される Llama 3.3 70B または Qwen 2.5 72B は、日常的なコーディングタスクのほとんどを処理できます。これにより、Cline はエアギャップや規制環境に最適な選択肢となります。
ユーザーレビュー
レビューを書く →Be the first to review. We verify every reviewer by email.
