Cline Cloud sessions keep your task running after you close VS Code

1 day ago

Cline Cloud sessions run your task in a hosted sandbox on a GitHub repo and branch you pick, and keep going after you close VS Code or JetBrains. Also in Cline v4.1.24 to v4.1.27: read_files sends whole files, LM Studio and Ollama get past the five-minute timeout, turning reasoning off works on GPT-6, broken streams are retried, .claude/skills is discovered, and with no saved model the Cline provider starts on a free one.

Ask about this presentation

Chapters

  1. 0:00Cline Cloud sessions
  2. 0:11Full file reads
  3. 0:18Local models
  4. 0:28Reasoning settings
  5. 0:38Stream retries
  6. 0:47Skills and rules
  7. 0:56Three fixes
  8. 1:05Default models
Show transcript
  1. Cline Cloud sessions run a task in a hosted sandbox, on a GitHub repo and branch you pick, and keep going after you close your editor. Cloud tasks get their own Cline model, and Cloud is rolling out gradually.

    Slide text
    Cline

    Cline Cloud runs your task after you close your editor

    A hosted sandbox on a GitHub repo and branch you pick · its own model, set in Settings

    cline/cline · v4.1.25 · v4.1.26 release notes
  2. When Cline reads a file, the model now gets all of it, up to two thousand lines. Before, file contents were cut to eight thousand characters on the way to the model.

    Slide text
    ClineCline
    Changed

    read_files hands the model the whole file

    Up to 2,000 lines or about 48k characters · no longer cut at 8k

    cline/cline · v4.1.27 release notes
  3. Long LM Studio and Ollama requests no longer die after five minutes of silence on a large prompt. LM Studio settings also take an API key now, for servers that require one.

    Slide text
    ClineCline
    Fixed

    LM Studio and Ollama get past the five-minute timeout

    Large prompts no longer end in BodyTimeoutError · LM Studio settings take an API key

    cline/cline · v4.1.25 · v4.1.27 release notes
  4. Turning reasoning off no longer fails with a 400 error on GPT-6, Claude Opus 5.5 or Gemini. GPT-6 also runs on OpenAI and Azure OpenAI compatible endpoints.

    Slide text
    ClineCline
    Fixed

    Turning reasoning off works on GPT-6, Claude and Gemini

    GPT-6 runs on OpenAI and Azure OpenAI compatible endpoints · “None” is saved and sent

    cline/cline · v4.1.24 · v4.1.26 · v4.1.27 release notes
  5. A provider error halfway through a reply used to fail the whole turn. Now Cline tries again, up to three times. Text that already streamed in can show up twice.

    Slide text
    ClineCline
    Fixed

    Cline retries a reply that breaks off mid-stream

    Up to 3 retries before the turn fails · text already streamed can repeat

    cline/cline · v4.1.25 · v4.1.26 release notes · PR #15100
  6. Skills you wrote for Claude Code, in a .claude/skills folder, now show up in Cline. Rule files that are symlinks load, and a rule you turn off stays off.

    Slide text
    ClineCline
    Fixed

    Cline finds the skills in .claude/skills

    Symlinked rule files load · disabled rules stay disabled

    cline/cline · v4.1.24 · v4.1.26 release notes
  7. A free-tier Gemini key stops failing every turn. Starting a command leaves your keyboard in the chat. And after you cancel a task, your next message no longer answers Cline's old question.

    Slide text
    ClineCline
    Fixed

    Three things Cline stops doing

    • Failing every turn on a free-tier Gemini API key
    • Sending what you type in chat to the terminal
    • Taking your next message as the answer to an old question
    cline/cline · v4.1.26 release notes
  8. With no saved model, the Cline provider now starts on a free one. I would update: the fixes need nothing from you. Cloud needs a GitHub repo, and it is still rolling out.

    Slide text
    Cline

    With no saved model, Cline starts on a free one

    Saved selections are untouched · other providers’ defaults are listed in Cline’s CHANGELOG.md

    cline/cline · v4.1.26 · v4.1.27 release notes

What happened in AI today, in your inbox

One short email each evening. Read it and you’re caught up.

Built with ❤️ in New York