Pane vs Vibe Kanban

the elephant in the room: it's no longer maintained

Bloop shut down on April 10, 2026. Vibe Kanban still runs, but nobody is paid to keep it working.

Bloop said it couldn't find a business model because most users were on the free plan, refunded paid subscriptions and handed the project to the community. The last release is v0.1.44 from April 24, 2026. A handful of community commits landed in September, but no new release followed, and an npm publish of 0.1.45 was withdrawn. Its security contact addresses bounce, and phone access depends on Bloop's relay servers, which still answered when we checked but have no one behind them.

Its founder said afterwards that the lab-made agent apps had become very good and that he wasn't excited by a roadmap of catching up with them. If you depend on Vibe Kanban, plan for it to drift as agents change, because each agent is pinned to an April 2026 version.

the real difference: a re-drawn chat or the real terminal

Vibe Kanban runs agents headless and draws its own view. Pane shows the agent exactly as it is.

Vibe Kanban talks to agents through their machine interfaces and shows each step as a conversation. That makes features like inline diff comments and a live preview possible, but you lose the agent's own interface: interactive slash commands don't work, some tools like Claude Code's questions weren't supported, and users report stuck subagents and missed prompts. It drives Claude Code with claude -p, which Anthropic had planned to move to a separate monthly credit in June 2026; that change is paused, so for now it counts against your plan like any Claude Code use.

Pane starts each agent's real command line tool in a real terminal. Every command, prompt and shortcut works the way the agent's makers built it, and Claude Code runs interactively on your normal plan.

what happened to the kanban

The board is the feature people chose Vibe Kanban for, and it no longer works out of the box.

The board turned issues into agent prompts and moved cards as pull requests opened and merged. Since the cloud move it needs a sign-in, and since v0.1.44 projects are export-only even for self-hosters, so people roll back to 0.1.43 and run the server themselves. What still works without an account is workspaces: a task, an agent and a worktree.

Pane never had a board. It has Sessions, an orchestrator chat that starts panes, hands out tasks and tracks pull requests, and the runpane CLI and MCP server for agents that plan and launch work.

windows and wsl

Pane is a native Windows app that also runs WSL repos inside the distro. Vibe Kanban works best run inside WSL with npx.

Vibe Kanban has an unsigned Windows installer and a list of Windows bugs, including installer loops. WSL isn't documented, but running it inside WSL with npx works well, which is how we tested it; worktrees then live in /var/tmp inside the distro.

Pane installs as a native Windows app on x64 and ARM64, and runs agents, terminals and git inside WSL for repos you keep there.

defaults

Every Vibe Kanban agent profile skips approvals by default, and telemetry turns back on when you sign in.

Vibe Kanban's default profiles run every agent with approvals off: Claude Code with --dangerously-skip-permissions, Codex with full access, Gemini and Qwen in yolo mode, and so on, with a warning on the first screen. Telemetry is on by default; you can turn it off, but signing in turns it back on, and error reports go to Sentry with no switch.

Pane starts each task in its own worktree, and its product analytics turn off in Settings → Privacy.

what we measured

Vibe Kanban is tiny: one server process using about 110 MB. Pane is a full desktop app.

Run with npx, Vibe Kanban fetches a 52 MB package and runs a single Rust server using about 110 MB of memory at idle, which is far lighter than any desktop app in this space, Pane included (780 MB). Its first start, including the download, took 37 seconds. From the first screen to a change merged into main took about 4 minutes 10 at a human pace.

how we tested

On October 3 and 4, 2026 we ran Vibe Kanban 0.1.44 with npx inside WSL2 (Ubuntu 24.04) and drove its web UI from a browser on Windows, and installed its Windows desktop app. We gave it and Pane 2.4.155 the same task in a private test repo, using Claude Code on the account's existing login, and followed it through to a merge. We read the source to confirm defaults like skipped approvals and telemetry. We backed up every agent config before the test and restored anything an app changed.

Facts about each product come from its own docs, release notes, issue tracker and source code, checked on the same dates. Measurements are single-machine numbers, not lab benchmarks.

What we ran

• → Pane 2.4.155 on Windows 11, native and WSL2 (Ubuntu 24.04) repos
• → Vibe Kanban 0.1.44 with npx inside WSL2, web UI in a browser; Windows desktop installer also installed
• → Claude Code (Opus 5.5) on the account's existing login

Who wrote this

We make Pane, so we have a stake in this. We tried to be fair: Vibe Kanban's strengths are listed as plainly as Pane's, every claim was checked against the product itself or its docs and source, and the screenshots are unedited captures from our own runs.

Corrections

If anything here is wrong or out of date, tell us and we'll fix it and update the date at the top. Open an issue on GitHub

Sources

• → Vibe Kanban on GitHub
• → Vibe Kanban docs
• → Vibe Kanban releases
• → Sunsetting discussion
• → Pane on GitHub
• → Pane on Windows and WSL