Cline 4.0 Migrates the VS Code Extension Onto the Shared Cline SDK, Adds ClinePass, a Customize Marketplace, Plugins, and Queued Prompts

Cline 4.0 Migrates the VS Code Extension Onto the Shared Cline SDK, Adds ClinePass, a Customize Marketplace, Plugins, and Queued Prompts

lschvn

Cline shipped v4.0.0 on June 26, 2026, a major version that jumps from v3.89.2 (June 11) and reframes what Cline is. Where the 3.x line was a VS Code extension that grew a CLI alongside it, 4.0 makes the shared Cline SDK the engine for everything and ports the VS Code extension onto it. Cline now describes itself as "the open source coding agent in your IDE and terminal," and the README is explicit that the SDK is "the same engine that runs the CLI, Kanban, VS Code extension, and JetBrains plugin." The extension rewrite is the headline; the subscription, marketplace, and plugin layers are what turn the rewrite into a platform bet.

The SDK migration: one engine across four surfaces

The core change in 4.0 is that the VS Code extension no longer runs its own legacy task implementation. Agent turns, tools, Plan/Act mode coordination, MCP, checkpoints, telemetry, provider changes, compaction, mistake limits, and task history now route through the shared SDK session layer. The extension build and package workflow also moves to Bun, which Cline already uses across the rest of the repo.

The architecture payoff is consistency. The Cline CLI (now on its own v3.0.x line, at cli-v3.0.31 as of June 27), the Kanban parallel-agent task board, the JetBrains plugin, and the VS Code extension all share one runtime, so a behavior fix or a tool change lands in every surface instead of being reimplemented per client. Plan/Act handling is reworked through SDK coordinators with closer CLI parity and automatic continuation when switching from Plan to Act. Terminal execution is simplified through the SDK run-commands path, with safer structured command formatting and clearer non-interactive guidance. The same consolidation logic is what drove OpenAI to build a governable platform around Codex 0.142 earlier this week, and the broader pattern of coding agents moving from single-surface tools to SDK-first platforms is now the default shape of the category.

ClinePass: a managed subscription inside the extension

The most user-visible product change is ClinePass. It is a built-in subscription and onboarding flow in the VS Code extension, covering provider selection, signup and subscription handoff, live model lists, entitlement and organization error states, out-of-credit prompts, and clearer ClinePass auth and error handling. Generic SDK provider settings and model-catalog support mean more providers can share the same model picker, reasoning controls, dynamic model IDs, and custom model handling.

Read plainly, Cline now ships a managed path: subscribe, pick a model, run. That is a shift for a project that built its audience on "bring your own API key," and it is the same shift the rest of the category has made as coding agents moved from hobbyist tooling to managed products. BYO-key providers are not removed; they sit alongside the managed flow, and the new providers.json plus shared model catalog is designed to preserve settings across provider switches.

Customize marketplace and Cline Plugins

The Customize marketplace consolidates Skills, MCP servers, and Plugins into one interface, with installed and marketplace tabs, search and filtering, install and uninstall flows, and enable and disable controls. It also supports plugin-bundled skills and can be reorganized under a Customize entry point.

Cline Plugins are the new extensibility primitive. They let you extend Cline with custom tools, workflows, skills, and MCP-powered capabilities tailored to a team or project, and they can be installed mid-session. MCP support now extends to plugins, and the MCP hub refreshes automatically after a marketplace install, so a newly installed server is available without a manual restart. The MCP Marketplace tab can also be disabled remotely while installed MCP servers remain accessible. This is the same extensibility vector that Claude Code, Codex, and the opencode desktop release have been converging on: a plugin and MCP layer that turns a coding agent into something teams can customize without forking it.

Queued prompts, provider rework, and safer defaults

A batch of developer-facing improvements rounds out the release. Queued prompts fix a long-standing annoyance: messages submitted while Cline is already working now queue, render while the current turn streams, and can be cancelled before they run, instead of being dropped or racing the active turn. Edit-and-regenerate lets you edit a previous user message and rerun from there, with Escape cancelling the edit locally and clearer Reset Chat and Reset Code labels.

Provider and model configuration is reworked around providers.json, the model catalog, and SDK session config, and the release wires in a broad set of providers and models: ClinePass models, Fireworks GLM 5.2, Kimi K2.6 Fast, Kimi K2.7 Code, Qwen 3.7 Plus, MiniMax M3, SAP AI Core, LiteLLM model fetching, Codex OAuth credentials, and OpenAI-compatible model settings. The shared generic settings components and reasoning selectors replace many provider-specific views.

Safety defaults tighten. Command auto-approval is now disabled by default for new and reset configurations, and the auto-approval UI is streamlined. The release also fixes SDK tool-result and provider-message budgeting by truncating large tool outputs by default, capping assistant text, limiting bash, file-read, and search output ingestion, bounding media budgets, and batching outdated-read rewrites to preserve provider prefix caches. Terminal reliability gets a long fix list, including standalone Windows output capture, hardened PowerShell handling, heredoc coalescing, and removal of duplicated command echoes.

What to watch

Two caveats are worth flagging before upgrading. Subagents are temporarily disabled in the VS Code extension while the SDK-backed experience stabilizes, so teams relying on multi-agent orchestration through Cline's subagent feature will need to wait for it to return. The legacy Explain Changes feature is removed as part of the SDK migration cleanup. Everything else is additive. The full changelog between v3.89.2 and v4.0.0 is on the compare view, and the CLI line continues independently on the v3.0.x branch. Cline has roughly 64,000 GitHub stars and sits in a category that is consolidating fast; 4.0 is the version where it stops being an extension and starts being a platform.

Frequently Asked Questions

OpenAI Codex 0.142 Adds Rollout Token Budgets, Multi-Agent Delegation, Indexed Web Search, and a Reorganized Plugin Marketplace

The OpenAI Codex 0.142 line shipped between June 22 and June 26, 2026 (rust-v0.142.0 through rust-v0.142.3 on github.com/openai/codex). The cycle turns the agent into something a team can govern: configurable rollout token budgets that track usage across threads and abort turns when exhausted, multi-agent delegation configurable as disabled, explicit-request-only, or proactive, an indexed web-search mode that permits live search while restricting page access to server-approved URLs, MCP tools that use tool search by default, a /plugins marketplace reorganized into OpenAI Curated, Workspace, and Shared with me sections, system proxy (PAC, WPAD, bypass) support on Windows and macOS for authentication, and a long list of remote-executor reliability and security fixes. It builds directly on the 0.141 Noise-encrypted relay and cross-OS PathUri layer from June 18.

Prettier 3.9 Overhauls Five Parsers: micromark for Markdown, yaml v2, GraphQL.js v17, a Rust-Based Flow Parser, and Angular

Prettier 3.9.0, released June 27, 2026 (prettier/prettier, blog post by Fisker Cheung), is a parser-heavy release that upgrades Markdown from remark-parse v8 to micromark v4 (better CommonMark and GFM compliance and a stack of long-standing bug fixes), YAML to yaml v2, GraphQL to GraphQL.js v17 (fragment arguments and directives on directive definitions), Flow to the Flow team's new Rust-based oxidized parser (roughly 37% faster on Prettier's valid Flow fixtures and 43% faster on flow_parser.js in local parser-only benchmarks), and Angular. The JavaScript and TypeScript printer is reworked too, particularly in --no-semi mode where comments around break and continue are now stable across repeated formats (an idempotency fix), plus redundant parenthesis removal in return statements, embedded-template interpolation alignment, and logical-not inlining. The release drops the legacy import ... assert {} syntax (Babel 8 removed the parser plugin; migrate to with), fixes a silently broken --cache-strategy content option, and stops EditorConfig files above Git worktrees from leaking in. The team reiterates pinning the exact version in package.json because the formatting changes will produce diffs.

Related articles

More coverage with overlapping topics and tags.

Claude Code Issue #74066: Users Report Cross-Workspace Context Bleed on Sonnet 5, Anthropic Has Not Yet Responded
ai

Claude Code Issue #74066: Users Report Cross-Workspace Context Bleed on Sonnet 5, Anthropic Has Not Yet Responded

An open bug filed against Claude Code on 2026-07-04 by an [Enterprise ZDR](https://docs.anthropic.com/en/docs/build-with-claude/zero-data-retention) user describes a working session on Sonnet 5 that suddenly starts referencing an unrelated Minecraft temple build, then doubles down on the wrong task in its recap. The reporter (GitHub: [@milesrichardson-edb](https://github.com/milesrichardson-edb), issue [anthropics/claude-code#74066](https://github.com/anthropics/claude-code/issues/74066)) is on Enterprise Zero Data Retention, the tier Anthropic specifically advertises as session-isolated. Triage on the reporter's local session JSONL at `~/.claude/projects/<encoded-cwd>/<session-id>.jsonl` finds the leaked text is not in the transcript, ruling out a local context bleed by file overlap. Four other users in the comments (with work histories going back to last year) describe near-identical behavior across Claude Code, Claude Mobile, and Claude deep research. The most plausible architectural fit is shared KV-cache state in inference ([per @yv3nne in the comments](https://github.com/anthropics/claude-code/issues/74066#issuecomment-4880448776)), but no Anthropic engineer has commented on the issue in the 22 hours since it was filed, and the issue reached the top of [Hacker News](https://news.ycombinator.com/item?id=42481789) on 2026-07-04. The tone in the thread is split: half suspecting a real platform cache reuse, half suspecting a [sonnet-5-specific hallucination triggered by a Pygments lexer](https://github.com/anthropics/claude-code/issues/74066#issuecomment-4880334711). Both readings are credible.
Claude Code 2.1.199 Adds Stacked Slash-Skill Invocations, Fixes Subagent Error Reporting, the Background-Agent Daemon Crash Loop on Linux, and the Streaming-Response Discard Bug
ai

Claude Code 2.1.199 Adds Stacked Slash-Skill Invocations, Fixes Subagent Error Reporting, the Background-Agent Daemon Crash Loop on Linux, and the Streaming-Response Discard Bug

Anthropic released Claude Code [v2.1.199](https://github.com/anthropics/claude-code/releases/tag/v2.1.199) on 2026-07-02, one day after the v2.1.198 feature release (Chrome GA, background-agent auto-PR, /dataviz, Gateway on AWS) and two days after v2.1.197 made Sonnet 5 the default model. v2.1.199 is a bug-fix release with 24 entries: one new feature (stacked slash-skill invocations load up to 5 leading skills in one prompt), and a long list of reliability fixes for subagent error reporting, the background-agent daemon on Linux, streaming-response preservation on mid-stream API errors, SSL/TLS error handling behind corporate proxies, plan mode tool-call handling, and a per-env retry policy that no longer caps the user-tunable retry count at 15. Most of the items are in the background-agent / subagent area, which is consistent with the v2.1.197 + v2.1.198 direction of treating delegation as the product surface for Sonnet 5.
Claude Code 2.1.198 Makes Claude in Chrome Generally Available, Lets Background Agents Commit, Push, and Open Draft PRs, Adds `/dataviz` Skill, and Brings Claude Platform on AWS to the Gateway
ai

Claude Code 2.1.198 Makes Claude in Chrome Generally Available, Lets Background Agents Commit, Push, and Open Draft PRs, Adds `/dataviz` Skill, and Brings Claude Platform on AWS to the Gateway

Anthropic released Claude Code [v2.1.198](https://github.com/anthropics/claude-code/releases/tag/v2.1.198) on 2026-07-01T20:45:36Z, the second consecutive day of Claude Code shipping after v2.1.197 (the Sonnet 5 default release) the day before. The headline change is that Claude in Chrome is now generally available inside Claude Code: the browser-side assistant that was a research preview moves to a stable channel and is no longer gated behind a feature flag or a waitlist. The release also lands the first end-to-end automation for background agents in `claude agents`, which now commit, push, and open a draft pull request when they finish code work in a worktree, instead of stopping to ask for permission. Other notable additions: a `/dataviz` skill for chart and dashboard design with a runnable color-palette validator, a Gateway upstream for Claude Platform on AWS (`anthropicAws`) with model-not-found responses advancing the failover chain, a built-in Explore agent that now inherits the main session's model (capped at opus) instead of running on Haiku, and subagents that inherit the parent session's extended thinking configuration. The release also removes the `/agents` wizard and replaces it with the conversational flow, fixes a long list of reliability bugs around background agents, network drops, and platform auth, and improves syntax highlighting by upgrading to highlight.js 11.

Comments

Log in Log in to join the conversation.

No comments yet. Be the first to share your thoughts.