Cline

• Import your history from Claude Code, Codex, and opencode. An Import button in the Sessions header (and a row in Settings → General) scans your local stores from all three tools and turns the conversations you pick into fully resumable Cline sessions. Sessions are grouped per tool with select-all and a search across title, folder, and first prompt; already-imported ones are shown as such so re-opening the dialog is safe. Imported sessions resume on your configured provider and model, not the source tool's. If you have history from any of these tools, onboarding now offers the import as a step
• Runs of a schedule now fold into a single collapsible sidebar row named after the schedule, with its run count, instead of one row per run all carrying the same prompt title. Expanding lists them newest-first as "Run N" with the usual status dot, time, hover card, context menu, and delete; the group holding the active session opens on its own
• Voice input now works on macOS. The app shipped without a microphone usage description or entitlement, so dictation failed silently
• Web search is now on by default

• Cline now handles a running Hub that is older than your CLI. Instead of quietly talking to a hub executing stale code, you get a prompt showing how many active sessions a replacement would interrupt, with enter-to-replace or escape-to-keep. The replacement drains the Hub first so in-flight turns finish, and a hub too old or wedged to accept the drain is left alone rather than killed
• Windows binaries are now Authenticode-signed via Azure Trusted Signing, and a launch blocked by application-control policy now prints an actionable error instead of failing bare
• Fixed the CLI dying when an enabled remote (SSE/streamable HTTP) MCP server is unreachable. The connect now has a 10s budget, so an offline server no longer stalls session startup past the Hub's deadline and tears the session down — previously the interactive TUI exited and one-shot runs failed
• Fixed tool calling being silently disabled for Dify, SAP AI Core, opencode, and Codex CLI models. Their catalog entries declare no capabilities, and the empty list was read as an authoritative denial that stripped every tool from the request
• Fixed images being dropped from file reads on models whose capability list is empty

• Fixed tool calling being silently disabled for gateway models whose catalog entry declares no capabilities. Three separate producers built gateway model definitions with their own hand-written capability translations and had drifted; the builtin-provider path emitted ["text"] where the others emitted undefined, and that list read as an authoritative denial that stripped every tool definition from requests to Dify, SAP AI Core, opencode, and the Codex CLI. One shared translator now serves every producer
• Fixed an empty capability list stripping image input. Two readers bypassed modelHasCapability and tested capabilities directly, where an empty list is not nullish but matches nothing: the file-read tool silently dropped every image from the request, and model pickers reported models as definitively lacking vision, attachments, and reasoning when nothing had been declared
• Langfuse tracing now works in minified release builds. initializeLangfuseTelemetry identified the OpenTelemetry provider by constructor name, which minification renames, so every production build silently initialized with readiness=false while unminified dev runs worked. Detection is now

• Marketplace is now a two-pane explorer: a browsable list on the left and full catalog metadata for the selected item on the right, with category tag filters that collapse behind a "more" toggle
• Stopping a session now actually stops everything it started. Stop stays available while child agents are running, and an abort propagates to delegated subagents and to teammates instead of leaving orphaned work running in the background; cancelled teammate tasks now persist as cancelled
• Fixed the ask-a-question tool's option text overflowing instead of wrapping
• You can now drop file attachments anywhere over the chat input, not just on the small attach target
• Cline provider models now refresh from the live catalog, so newly released models show up without waiting for an app update
• Provider 401/403 responses are now classified as authentication errors rather than generic request failures, so a bad or missing API key is distinguishable from a real provider outage
• Fixed Langfuse tracing never initializing in release builds — the minified bundle broke tracer detection, so telemetry worked in dev and silently did nothing in the shipped app. Also updated for AI SDK 7's telemetry API

• Beta: hand off local sessions to Cline Cloud and continue working from cloud workspaces, with recovery for interrupted transfers and preservation of the prompt, attachments, and session state.
• Beta: choose between local, SSH remote, and Cloud environments from the desktop app, with the experimental realtime voice and avatar overlay experiences included.
• Beta: the GitHub onboarding step is available behind the code-onboarding-github feature flag and remains disabled by default.
• Includes all stable desktop improvements through 0.0.20, including the Windows release, full-history session search, scheduled-task fixes, inline tool-result images, and the latest provider and Marketplace updates.

Full Changelog: https://github.com/cline/cline/compare/desktop-v0.0.21-beta.1...desktop-v0.0.21-beta.2

• Cline Desktop now ships on Windows: releases include a code-signed x64 installer, and installed apps auto-update on the same feed macOS does
• Windows shell fixes: background processes (the sidecar, git) no longer pop visible console windows; updates now download in the background and install when you restart the app; the MCP settings path falls back to USERPROFILE when HOME is unset
• Tool results that return images — screenshots from browser or MCP tools — now render as inline images you can click to expand, with a carousel for stepping through multiple images, instead of raw base64 text
• Session search now covers your full indexed history. The sidebar search icon opens the command bar (Cmd/Ctrl+P) with server-ranked results, instead of a sidebar-local dialog that first loaded every session into memory
• Onboarding has a new GitHub integration step
• Fixed scheduled tasks disappearing after the app updated — hub-managed schedules were being wiped by cron reconciliation on restart
• Agent-created schedules now live in one user-level home (~/.cline/schedules) instead of being scattered across whichever chat folder created them, and they now appear on the Schedules page

• Fixed the background hub process ballooning in memory during long sessions — session status updates were broadcasting a full copy of the conversation transcript to every connected client, which on a large task could grow the process to tens of gigabytes. Upgrading retires the running hub so the fix takes effect on the next command
• New files are now created with your platform's native line endings
• Fixed the codebase search tool crashing on files that contain a single enormous line
• Cost estimates are no longer shown for Claude Code. Its usage is typically covered by a Claude Pro/Max subscription, but its models reuse Anthropic API pricing, so Cline was showing charges you were not being billed
• Credentials embedded in git remote URLs are now redacted from the workspace information sent to the model
• Installing an MCP server no longer misreads a -- separator in the install arguments as part of the server command
• Refreshed the model catalog. Adds seven providers (Agnes AI, Aixy, IteraCompute, LLM Tech, NeoSmith, Pendra, and Standard Compute) and updates model lists and pricing across providers. The resolved default model changes for ClinePass (now GLM 5.3), Z.ai, Hugging

• Session snapshot events no longer carry the full conversation transcript. Every session.updated (and session.created / session.detached / run.started) event embedded the session's entire message history, so on a multi-megabyte task each status flip shipped megabytes to every subscriber, flooded the durable event log, and could grow the hub process by one transcript copy per event — reported as a 25 GB cline process on a 16 GB machine. Snapshots are now state-only (status, usage, model, workspace, checkpoint); transcripts are fetched with the session.messages command. Checkpoint-restore replies, which carry messages in their own field, are unaffected

Full Changelog: https://github.com/cline/cline/compare/sdk/sdk/v0.0.80...sdk/sdk/v0.0.81

• Fixed the background Cline process ballooning in memory during long sessions — session status updates were carrying a full copy of the conversation transcript to every connected client, which on a multi-megabyte task could grow the process to tens of gigabytes. Status updates now carry only state (status, usage, model, workspace, checkpoint); the transcript is fetched on demand
• Refreshed the model catalog. Adds seven providers (Agnes AI, Aixy, IteraCompute, LLM Tech, NeoSmith, Pendra, and Standard Compute) and updates model lists and pricing across providers. The resolved default model changes for ClinePass (now GLM 5.3), Z.ai, Hugging Face, evroc, LLM Gateway, NanoGPT, and Weights & Biases, so if you use one of those without pinning a model you will get a different default

Full Changelog: https://github.com/cline/cline/compare/desktop-v0.0.18...desktop-v0.0.19

• New files created by the file-writing tools now use the platform's native line endings
• Fixed search_codebase crashing the process on files that contain a single enormous line
• Credentials embedded in git remote URLs are now redacted from the workspace information included in the system prompt
• Claude Code is now marked as a subscription-billed provider, so hosts suppress its per-token prices and API-rate cost estimates. Its models reuse Anthropic API pricing metadata, which produced dollar figures for usage that is covered by a Claude Pro/Max subscription
• installMcpServer no longer treats a -- separator in the install arguments as part of the stdio command
• The tasks tool can now be configured to offer only scheduled work, disabling the Todo kind while leaving the Agenda backend intact
• Refreshed the model catalog. Adds seven providers (Agnes AI, Aixy, IteraCompute, LLM Tech, NeoSmith, Pendra, and Standard Compute) and updates model lists and pricing across providers. The resolved default model changes for ClinePass (now GLM 5.3), Z.ai, Hugging Face, evroc, LLM Gateway, NanoGPT, and Weights & Biases, so if you use one of those without pinning a model you will get

Everything here lands through the SDK bundle, so it applies to windows running that bundle.

Fixed

• Cost estimates are no longer shown for providers billed by a flat-rate subscription (ClinePass, ChatGPT via Codex, and Claude Code). The task header and model pricing rows rendered API-rate dollar figures that read as real charges on top of the subscription, including a flash of them on every chat-view mount while provider listings were loading.
• Signing back in no longer moves your last-used provider off ClinePass on credential refresh.
• Hooks now resolve their workspace from the VS Code window instead of shared global state in ~/.cline. With a second window open on another project, a workspace's .clinerules/hooks scripts were never discovered, and hook cwd and the workspace paths passed to hook scripts resolved against whatever project some other or older Cline instance last recorded.
• New files are now created with your platform's native line endings.
• Fixed the codebase search tool crashing on files containing a single enormous line.

• The sidebar is time-sorted again by default, with collapsible Pinned / Scheduled / Tasks sections and a one-click toggle to switch to project grouping (the old dropdown is gone). Scheduled sessions are marked with a clock icon, and the list starts taller and grows to fill the sidebar instead of stranding rows over empty space
• Session rows now show a trash button on hover for quick deletion, with the same confirmation the row's context menu uses
• Customize is now your installed inventory only. Browsing moved to a dedicated Marketplace page — one list across plugins, MCP servers, and skills with type-filter and tag chips — and the two pages link to each other from their headers and from sidebar sub-tabs
• Schedule cards are now click targets: clicking a card anywhere outside its controls opens its details, the redundant eye button is gone, and the edit / run / pause / delete buttons are large enough to hit
• Schedule details are one scrollable view instead of Overview/Runs tabs, showing the meta grid, the configuration, and the most recent runs with a "Show all N runs" expander
• "Run now" now hands you into the session it starts

• Plugins, MCP, Skills, Rules, Hooks, and Tools are now one Customize hub with tabbed sections and live counts. Catalog-backed tabs show what you have installed followed by an inline Browse section, so installing something from the catalog immediately appears above — the separate Marketplace page is gone
• Redesigned the Models page: providers are grouped into Connected, Popular, and All with their auth kind and configuration status instead of per-row toggles. OAuth providers now offer a browser sign-in rather than an API key field, with a collapsed manual-key escape hatch where supported, and explicit Connect / Disconnect / Sign out actions
• Voice input moved to its own Settings → Voice page that only offers connected transcription-capable providers and preselects a default model. The composer's microphone button now appears only once a voice model is configured
• Sidebar sessions are always grouped by project, with pinned sessions leading each group and scheduled sessions marked by an inline clock. The Favorite action is now called Pin

• The first-launch "Try ClinePass" dialog no longer advertises the $4.99 first-month promo, which is ending
• The hub's event log is now capped at 64 MiB on disk. Events carrying full session snapshots could previously grow the log to tens of gigabytes on a long-running hub, since deleting rows never shrinks the file. Oldest events are dropped first and the space is returned, and pruning runs on volume as well as on a timer
• Refreshed the model catalog. Adds two providers (AgentRouter and Opper) and updates model lists and pricing across providers. The resolved default model changes for Aki.io and NanoGPT, so if you use one of those without pinning a model you will get a different default

Full Changelog: https://github.com/cline/cline/compare/cli-v3.0.57...cli-v3.0.58