CLI v3.0.63

• Context compaction no longer silently degrades to truncation partway through a long session on OAuth providers. The summarizer kept the access token resolved when the session started while the main agent loop refreshed its own each turn, so once the token rotated the summarizer's request failed with 401 and the failure was swallowed by the truncation fallback — you got a chopped transcript instead of a summary, most visibly on cline-free/* models. The summarizer also stayed on the original model after a mid-session model switch. Both now follow the session's current credentials and model
• Startup no longer waits on a feature-flag network round trip. The CLI awaited a PostHog poll before the session was ready; flags now refresh in the background, and a concurrent poll no longer reads a stale default while the first request is still in flight
• Context injected by a hook is no longer mistaken for something you typed. The prompt_submit dispatch and the audit logger both fired on injected hook-context messages, which carry the user role but a system display role. Run-start hook control (cancel and context injection from agent_start/agent_resume scripts) remains inert in the CLI pending a deliberate opt-in
• Compaction now fires on your provider's actual token usage instead of a character estimate. Dense content — disassembly, image dumps, minified sources — tokenizes far denser than the estimate assumed, so a long session could hit the real context ceiling without ever compacting and then get squeezed down to a handful of output tokens per turn. The summarizer also gets a larger output budget, since a model that reasons by default could spend a tight one thinking and return no summary at all
• Subagents now run their tool calls concurrently, instead of the first child finishing before the next starts. Approvals and hooks still run in order, and the parent still waits for every result before its next turn
• Configured subagents no longer ask you to approve their own tool calls after you have already approved the delegation. With concurrent siblings this was also unsafe: several children could reach the terminal prompt on the same stdin at once, where a single y could approve more than one pending operation
• Deleting a session from history now actually removes it. A session whose index row was missing but whose on-disk manifest survived was reported as deleted and then reappeared on the next refresh
• .cline/rules and global rules are now found consistently. Only one of the two supported workspace rule layouts was being looked for, and the CLI could miss global rules that the VS Code extension had created in a OneDrive-redirected Documents folder
• Press Enter on an empty prompt to steer the first queued message to the front of the queue
• Typing an answer to an ask_question prompt now keeps your capitalization, punctuation, spaces, and Unicode. The input reconstructed text from the normalized key name, so shifted letters came out lowercased, and backspace could leave malformed text behind after an emoji or combining mark. Backspace now deletes one whole grapheme
• On Windows, a program planted in your workspace can no longer run in place of the real one. Windows resolves a bare program name by searching the working directory before PATH, and the CLI spawns rg, git and powershell with your repo as the working directory, so opening a repo containing an rg.exe ran it during file indexing, before any approval. The CLI and the hub daemon it boots now opt out at startup
• Non-interactive commands no longer load the interactive runtime. Commands like cline history --json pulled in core through the MCP and schedule wizards; on Windows an isolated history listing went from 6–7.5 seconds to about 0.85 seconds
• Login-shell PATH probing waits up to 5 seconds per shell instead of 2. nvm-style profiles regularly need longer than 2 seconds on older machines, and a failed probe leaves the runtimes Cline spawns missing from PATH
• CoreWeave replaces the Weights & Biases branding and docs link on that provider. Your provider ID and stored credentials are unchanged
• Refreshed the model catalog, from 203 to 209 providers and 6,079 to 6,188 models. Kimi For Coding splits into separate kimi.com and kimi.ai providers, and AI21 Labs, ainetcafe, Inco, OCI Generative AI, Tempr, and Vispark are new. The resolved default model changes for 29 providers — eight land on DeepSeek V4.1 Flash (Alibaba CN and its token plan, Nebius, Neuralwatt, SCNet, Synthetic, TensorX, Vancine) and six on GLM 5.3 Flash or FlashX (iteracompute, Privatemode AI, SiliconFlow, Volcengine, ZenMux, Zhipu AI), while GitHub Copilot moves from GPT-6 Astra to Grok 4.7 and StepFun to Step 5 Preview. If you use one of those providers without pinning a model, expect a different default

Full Changelog: cli-v3.0.62...cli-v3.0.63