C

Cline

client Website ↗ Website ↗ Verified 2026-11-17

📊 Scores

Coding Ability
8.2
Value
9.6
Flexibility
9.5
China Access
9.0

🚀 Quick Start

Install Steps

  1. 1 Open VS Code, search for "Cline" in the extension marketplace and install
  2. 2 Click the Cline icon in the sidebar, select model provider and enter API Key
  3. 3 Configure the model using current provider documentation; do not rely on expired introductory promotions or direct-China-access promises
  4. 4 Describe tasks in chat; Cline reads/edits files and runs terminal commands automatically
Register / Get Account ↗
🇨🇳 China Access Note: Domestic APIs or local models are options, but provider regional policies, reachability and charges still apply; no mainland network test was performed.

💲 Pricing

Plan Price Limits
开源免费 $0 Individual open-source VS Code/CLI client is free; BYOK or Cline-provider inference costs are separate Official →

Open-source code is Apache-2.0. JetBrains plugins are not currently open-sourced and are listed under custom-priced Enterprise; not every client is free/open source. Cloud, ClinePass, SDK and extension have separate version/commercial scopes; v4.1.26 here tracks the extension. Auto-approval and Cloud sessions differ in permission behavior; context goes to the selected model provider. Historical editorial scores retained, not newly measured or re-rated; not a fresh performance or cost benchmark.

Cline

Free software/tier listed; check pricing scope and signup terms

Website ↗

✨ Key Features

Code Completion —
Chat ✅
Agent Mode ✅
Multi-file Edit ✅
MCP Support ✅
Context Window 取决于所选模型
Supported Models Anthropic、OpenAI、Gemini、OpenRouter、兼容 API 或 Ollama / LM Studio;可用模型受提供商和功能限制
ide_base VS Code 插件;另有 CLI / macOS、Windows 桌面端 / Enterprise JetBrains
cli_support ✅
supported_os macOS, Windows, Linux
jetbrains_support ✅
Git Integration ✅
plan_act_mode ✅
auto_approve ✅
browser_preview ✅

🎯 Best For

Developers who want full model freedomMulti-provider users willing to compare model charges and plan restrictionsVS Code users who want Agent capabilities

📝 Changelog

2026-10-26 v4.1.26: with cloud sessions on, Settings > API Configuration has a Local | Cloud toggle. Cloud stores the Cline model used by new cloud tasks separately from local models; clicking the model label in a cloud session composer opens that picker. Starting or resuming a cloud sandbox shows a progress card with the repo, branch, and steps (start sandbox, clone repository, start agent), then collapses once the agent starts. Cloud sessions hide the Plan/Act toggle and the auto-approve bar. With no model chosen, the Cline provider uses Claude Sonnet 5.5 again, back at the top of the recommended list. Model requests retry at most 3 times, down from 5. Kimi K3 Extra High and Max now send max reasoning instead of an xhigh value the model rejects. The catalog adds CommandCode, Reflection, Liquid Inference, and Munito AI, and drops CrofAI, iFlow, and Llama. The Cline free list adds Reflection Beam, Nemotron 3.5 Lightning, and Laguna S 2.1, and drops Mimo V2.6 Flash and Muse Spark 1.3 Contributor. ClinePass now defaults to Kimi K3.1 Pro and adds MiniMax M3.1 Flash and DeepSeek V4.1 Flash 1017. Default models change for most other providers, often to those three; an unpinned model will follow the new default. Fixes: GPT-6 models such as gpt-6-sol and gpt-6.1-sol work on OpenAI and Azure OpenAI-compatible endpoints, and SAP AI Core no longer rejects max_tokens; both send max_completion_tokens. Gemini API keys on the gemini provider no longer send native Google Search; free-tier keys failed every turn with HTTP 429 on Gemini 3+, and on Gemini 2.x every Cline tool was dropped. Toggling Plan/Act while a task is starting no longer drops the first prompt or keeps the other mode tools. Disabled rules stay disabled. Starting a command no longer moves keyboard focus into the terminal. Cancelling a task, or opening another task while Cline is asking a question, no longer treats the next message as the answer to the old question. A turn that drops after the model has streamed only reasoning is retried instead of ending the task. Auto-compaction no longer runs for models with an unknown context limit, and it no longer runs again every few tool calls after the first compaction. ask_question continues after an option is picked, and empty answers are rejected. run_commands with an empty command list returns an error instead of a silent success; on PowerShell it explains quote escaping and asks the model to confirm recursive deletes first. Task header badges wrap so the close button stays visible. Verified 2026-11-03 from the GitHub Release as the latest stable extension release (v-prefixed; cli, sdk, and desktop lines excluded).
2026-10-16 v4.1.25: Cline Cloud sessions roll out gradually, running a task in a hosted sandbox on a chosen GitHub repo and branch (enable in Settings). Cloud tasks appear in Recent and History, notify on completion, and keep running after the editor closes. With no model pinned, the Cline provider defaults to Gemini 4 Argon (first on the recommended list); OpenRouter stays on Claude Sonnet 5.5. Catalog adds Boundless and Standard Compute; SambaNova now comes from the catalog with MiniMax M3 as its default; the recommended list adds Gemini 4 Argon and drops Grok 4.7; the free list replaces Step 5 Preview with Nemotron 3.5 Super VL; several provider defaults change. Fixes local-model requests (LM Studio, Ollama) failing after 5 minutes of silence with BodyTimeoutError, host-only OpenAI Compatible base URLs silently gaining /v1, attachments on answers to Cline questions being dropped, mid-stream provider errors not retrying, the JetBrains Run task card sitting under the Auto-approve bar, and Auto Compact separating from its strategy setting. Verified 2026-10-22 from tags as the latest stable extension release (v-prefixed; cli/sdk/desktop lines excluded).
2026-10-13 v4.1.24: Cline and OpenRouter default to Claude Sonnet 5.5 when no model is pinned; ai& gets its own API key and model picker; catalog adds Claude Haiku 5.5 plus AI/ML API, LowRouter, and Prime Inference. Fixes 400s when reasoning is off for GPT-6 Astra, GPT-6.1 Sol, Claude Fable 5, and Claude Opus 5.5, extension-host crashes when an MCP server or hook exits early, stale-model 404s shown as signed-out, images inflating token estimates and triggering early compaction, and loading of symlinked rules plus .claude/skills. Verified 2026-10-16 as the latest stable extension release (v-prefixed; cli/sdk/desktop lines excluded).
2026-10-07 v4.1.23: continues once when a response ends without a recognized finish reason; Generate Commit Message follows .clinerules; custom Anthropic base URLs (Azure AI Foundry, corporate gateways) no longer 400 from an option added in 4.1.22; oversized MCP output can be paged through; tool approve/reject works while the panel is still updating.
2026-09-30 v4.1.22: adds Bee (by HEOSSI) and Pareto Inference. Direct Anthropic calls use server-side refusal fallback, and Anthropic models on OpenRouter and Cline can fail over to another upstream. Content-filter blocks are explained instead of shown as an empty response. Recommended list adds Claude Sonnet 5.5 and Claude Opus 5.5; several providers default to GPT-6.1 Sol.

📦 Plans Using This Tool

📖 Related Articles

📚 Related Guides

🔗 Related Tools

💬 Comments

Last Verified: 2026-11-17 Next Review: 2026-12-17 Confidence: high Report Error →