Cline Review 2026: Pricing, Features and Verdict

What it is

Cline is an open source AI coding agent from Cline Bot Inc., licensed Apache 2.0, with about 70,500 stars on GitHub. It started as a VS Code extension (the marketplace ID is still saoudrizwan.claude-dev) and now runs in four places:

• VS Code extension, the core product, with about 5.7 million installs.
• Cline CLI for terminal workflows and scripts.
• Cline for Desktop (beta), a native app for macOS, Windows and Linux that opens a workspace without an editor. It was at version 0.0.48 in October 2026, and Cline warns to "expect rough edges."
• Cline Cloud, a browser mode that runs tasks in cloud sandboxes and pushes to a branch.

Cline is model agnostic. It does not sell a subscription bundle; it is a harness you point at whichever model you choose.

Pricing explained

Plan	Price	Includes
Open Source	$0	VS Code extension, CLI, client side architecture, inference at cost or BYOK, MCP marketplace, multi root workspaces, community support
Enterprise	Custom	JetBrains extension, SSO, SLA, dedicated support, centralized billing, RBAC, limits on inference providers, team dashboard, authentication logs

As listed on cline.bot/pricing in October 2026.

What you actually pay for. Model inference, billed by your provider per token. Cline lists support for Anthropic, Gemini, OpenAI, OpenRouter, AWS Bedrock, GCP Vertex, Groq, Cerebras, Vercel AI Gateway, DeepSeek and more, and you can also buy inference at cost through the Cline provider.

Illustrative monthly bills at Anthropic's listed API rates, for 20 million input and 2 million output tokens (a busy month of agent work, before caching):

Model	Input per million	Output per million	Month
Claude Haiku 5.5	$0.10	$0.50	$3
Claude Sonnet 5.5	$2	$10	$60
Claude Opus 5.5	$4	$20	$120

Agents resend context on every turn, so input tokens dominate. Prompt caching reads cost a fraction of the input price ($0.10 per million on Sonnet 5.5), which can cut the input side sharply for long sessions.

How that compares. A flat $20 plan (Claude Code, Codex, Windsurf) or $10 GitHub Copilot beats Cline for heavy frontier model users. Cline wins for light users, for people who route routine work to cheap models, and for teams that already buy models through AWS or Google Cloud. See the Price Index.

Key features

Plan and Act modes

Cline separates planning from doing. In Plan mode it reads the codebase and proposes an approach; in Act mode it edits files and runs commands, asking for approval at each step unless you allow auto approval (YOLO mode). This is the main reason people trust it with real repos.

Any model, switch any time

Pick a model per task and change it mid session, from local and open weight models to frontier ones. That makes it easy to plan on a strong model and implement on a cheap one.

MCP marketplace

A built in marketplace for Model Context Protocol servers adds tools such as databases, browsers and issue trackers without writing config by hand.

Cline Cloud

Connect GitHub, pick a repo and a branch, describe the task and walk away. Each session runs in its own sandbox and commits to a cline/ branch as it goes. You can run up to 10 sessions at once, follow progress on mobile, paste screenshots, and the GitHub token is scoped to the single repo you chose. Free models are available to start.

Desktop app (beta)

Queue follow up messages while the agent works, fork and edit an earlier message to rewind the workspace to that checkpoint, and see sessions from the desktop app, CLI, extension and scheduled runs in one history.

Multi root workspaces and SDK

Work across several folders at once, and build on Cline's SDK for custom workflows.

What users say

• VS Code Marketplace: 4.0 from 323 reviews, with about 5.7 million installs. That puts it level with GitHub Copilot (4.1 from 1,058) and above Claude Code (3.7 from 829) and Codex (2.9 from 722) on the same marketplace.
• Trustpilot shows only 3 reviews, too few to rate. Those three complain about usage running out fast and the agent making edits the user did not want, which are the two risks to manage with any agent: watch spend and keep approvals on for risky commands.
• What the marketplace numbers suggest: a large install base and an above average rating for a tool with no paid tier, which fits its pitch of control and no lock in.

Privacy and security

Cline calls its design a secure client side architecture: the extension runs on your machine and sends requests to the model provider you configure. That means the data retention and training terms that apply are your provider's, not Cline's, which is an advantage if your company already has an agreement with AWS Bedrock, Google Vertex or a model lab. The code is open source, so security teams can audit it. Enterprise adds SSO, OIDC, SCIM, RBAC, the ability to restrict which inference providers developers can use, authentication logs, VPC deployment and OpenTelemetry. Cline Cloud runs each session in an isolated sandbox with a repo scoped token.

Who it is for

• Developers who want to bring their own API keys and keep model choice.
• Companies with existing Bedrock or Vertex contracts that want an agent on top.
• Cost conscious light users, who pay a few dollars a month on small models.
• Teams that need open source for audit or self hosting reasons.

Who should skip it

• Anyone who wants a capped bill. There is no plan ceiling; set spend limits with your provider.
• JetBrains users on a personal budget; JetBrains support is Enterprise only. JetBrains AI (Junie) or GitHub Copilot cover those IDEs.
• Heavy daily users of frontier models, who will usually pay less on a $20 to $200 flat plan.
• People who want a polished, all in one editor; see Cursor or Windsurf.

Alternatives

• GitHub Copilot: $10 flat with credits for four labs' models.
• Claude Code: Claude usage included from $20.
• Kiro: published credits at $0.04 each, with open weight models.
• Cursor: full editor, third party models at API list price.

Verdict

Cline is the best free and open source option among the agents we compared: no subscription, any model, clear Plan and Act control, and now a CLI, a desktop beta and a cloud mode. Its 4.0 rating from 323 VS Code reviews is among the best in the category. It ranks below the flat fee leaders only because cost is open ended and team features sit behind Enterprise. If you want full control and are willing to watch your token spend, start here.