Do I need a Pane scheduler to run loops?

No. Run a loop from Codex, Claude Code, a shell script, cron or any orchestrator. Pane gives it the workspace commands it needs: repos, panes, terminal output and input. Pane Chat and Sessions are built-in loop agents if you'd rather not set one up.

Is this only for Codex or Claude Code?

No. Pane runs any agent that works in a terminal. Claude Code, Codex and Cursor have built-in presets, and anything else runs as a custom command.

Does this work on Windows and WSL?

Yes. Pane runs on Windows, macOS and Linux. If your loop agent runs inside WSL while Pane is a Windows app, call runpane through PowerShell, as shown below.

How do loops avoid burning context?

runpane keeps output short by default: the command summary is brief, terminal reads are limited, and full details load only when asked for. runpane watch can also batch routine events so the loop agent isn't woken for every line.

Yes, AGPL-3.0. The app, the runpane CLI and the orchestrator skills are all public.

Agent loops

how to set up your first ai agent loop

TL;DR

An agent loop is one agent that runs your workspace: it picks up work, creates a pane per feature, opens review and QA tabs in it, checks their output, answers their prompts and tells you what needs review. Pane gives that agent what it needs through the runpane CLI: saved repos, worktree-backed panes, terminals it can drive, an event stream, and short output reads that don't flood its context. It works with Codex, Claude Code, Aider, Goose or any CLI, on Windows, WSL, macOS and Linux. This page has a first-loop recipe and a starter prompt.

Instead of babysitting terminals, you have one agent that manages the workspace. It picks up work, creates a pane per feature, opens review and QA tabs in it, checks their output, sends follow-up input and tells you what needs review. Pane makes that work with any agent, on any OS.

Any agent

Codex, Claude Code, Aider, Goose, OpenCode or any terminal CLI.

Any OS

Windows, WSL, macOS, Linux, VMs and remote hosts.

Open source

AGPL-3.0. The app, CLI and orchestrator skills are all public.

The shape of a loop

Not a magic prompt: a small operating cycle of inspect, spawn, drive, watch, validate, report.

Step	What the agent does	What Pane gives it
Inspect	Read issues, pull requests, CI or repo state	Saved repositories and existing panes
Spawn	Create a pane for each feature and tabs for its reviews and QA	Worktree-backed panes and agent tabs
Drive	Start agents, answer prompts, send input	Terminal-backed panels
Watch	Wake up when an agent finishes or gets stuck	An event stream: ready, blocked, stuck
Validate	Read recent output before claiming success	Short, limited terminal reads
Report	Summarize what needs human review	Visible panes, diffs, branches and statuses

First loop recipe

Install, discover, add the repo, create a pane per feature, check, continue.

1Install or update Pane
Use the package-manager path so you and your agents rely on the same runpane command. Inside Pane's own terminals, runpane is already on the PATH.
```
npx --yes runpane@latest
pnpm dlx runpane@latest
pipx run runpane
```

2Let the agent find Pane

Have the agent check what commands exist. The summary is short by default; full details for one command load only when needed.

curl -fsSL https://runpane.com/runpane-cli-contract.json
runpane doctor --json
runpane agent-context --json
runpane agent-context --command "panes create" --json

3Add the repo if needed

Don't assume Pane knows the repo. Check, then add it if it's missing.

runpane repos list --json
runpane repos add --path "$PWD" --yes --json
runpane agents doctor --agent codex --repo active --json

4One pane per feature
1 feature = 1 worktree = 1 branch = 1 Pane; reviews, audits, fixes and QA for it are tabs in that Pane. A pane is a git worktree, and creating one installs dependencies and often builds, so create one only when the work's branch has no pane yet. To find the pane for a PR, get its head branch, then check each pane's worktreePath from panes list:
```
gh pr view <number> --json headRefName
runpane panes list --repo active --json
git -C <worktreePath> branch --show-current
```
If no pane owns the branch, create one. Use a built-in agent or pass any terminal command.
```
runpane panes create \
  --repo active \
  --name "issue-51-login" \
  --agent codex \
  --prompt "Implement issue #51. Open a PR when checks pass." \
  --source agent \
  --wait-ready \
  --yes \
  --json
```
Every later agent for that feature opens as a new tab in its pane. Tabs share one checkout: only the implementer moves HEAD, and review, audit and QA tabs never run git checkout, reset or stash.
```
runpane panels create \
  --pane <pane-id> \
  --agent codex \
  --initial-input "Review the PR from this branch. Run checks, summarize findings, and stop before merging." \
  --source agent \
  --no-focus \
  --wait-ready \
  --yes \
  --json
```

5Check what actually happened

After creating a pane or tab, or sending input, verify it. screen shows what's on screen now; output gives scrollback; watch streams ready, blocked and stuck events so the loop wakes only when needed.

runpane panes list --repo active --json
runpane panels wait --panel <panel-id> --for ready --timeout-ms 30000 --json
runpane panels screen --panel <panel-id> --limit 80 --json
runpane panels output --panel <panel-id> --limit 200 --json
runpane watch --follow --json

6Answer prompts and continue

Use panels submit for ordinary follow-ups, panels submit-composer for Codex or Claude composer boxes, and panels input only for exact bytes such as Ctrl-C.

runpane panels submit --panel <panel-id> --text "Continue with the review pass." --yes --json
runpane panels submit-composer --panel <panel-id> --strategy auto --yes --json
printf '\003' | runpane panels input --panel <panel-id> --input-file - --yes --json
runpane panels wait --panel <panel-id> --for idle --timeout-ms 300000 --json

A starter prompt for your orchestrator

Give any shell-capable agent this prompt and it becomes your loop runner.

Give this to Codex, Claude Code or any agent that can run shell commands:

You are running an agent loop for this repo.
Every cycle:
1. Inspect GitHub issues and PRs that need action.
2. Fetch the public contract if needed, then run runpane doctor --json and runpane agent-context.
3. Make sure this repository is saved in Pane and the requested agent is runnable.
4. 1 feature = 1 worktree = 1 branch = 1 Pane; reviews, audits, fixes and QA for it are tabs in that Pane. Create a Pane only when the work's branch has no Pane yet. Find an existing Pane with runpane panes list --repo <repo> --json by matching the PR's head branch against git -C <worktreePath> branch --show-current.
5. Start the right terminal agent with a clear initial prompt. For a new Pane use runpane panes create --source agent --wait-ready. For a tab use runpane panels create --pane <id> --agent <agent> --initial-input "<prompt>" --source agent --no-focus --wait-ready --yes --json. Tabs share one checkout: only the implementer moves HEAD, and review, audit and QA tabs never git checkout, reset or stash.
6. Use panels wait and panels screen before claiming a pane or tab started correctly.
7. Send follow-up input with panels submit or submit-composer; use panels input only for exact bytes.
8. Report the panes and tabs created, what each is doing, and what needs human review.
Keep output brief. Do not merge without explicit human approval.

Or use a built-in loop

Pane Chat and Sessions are loop agents that ship with Pane.

If you don't want to set up your own runner, use Pane's. Pane Chat is a chat with Claude Code, Codex or Cursor that already knows the runpane commands and the orchestrator skill. Sessions are named, persistent versions with their own workspace. Both stop and ask before merging or deploying. Ask for unattended mode to let them resume a pane whose turn died from a sleep or network error.

Windows and WSL

From inside WSL, reach the Windows Pane through PowerShell.

Pane can run as a Windows app while your loop agent lives inside WSL. A Linux shell may not see the Windows Pane directly, so call runpane through PowerShell and pick the saved repo by name or id:

powershell.exe -NoProfile -Command \
  'Set-Location $env:TEMP; runpane repos list --json'
powershell.exe -NoProfile -Command \
  'Set-Location $env:TEMP; runpane panes create --repo "Pane" --name "loop-smoke" --agent codex --prompt "Say hello from this loop and wait." --source agent --wait-ready --yes --json'

Pane handles the workspace, not the world

Your agent still uses GitHub, MCP and trackers directly.

Pane gives your loop the workspace commands. Your agent still uses the GitHub CLI, MCP servers, Linear, Jira or anything else for the outside world, so you're not locked into one model, editor or host.

来源

勘误

如果这里有任何错误或过时的内容，请告诉我们，我们会修正并更新顶部的日期。在 GitHub 上提交 issue

来源

常见问题

或

下载*

Windows SmartScreen 警告Pane 尚未签名，直接下载时可能会出现 SmartScreen 警告。 Pane 完全开源，你可以审查代码，也可以自己从源码构建。1. 点击“更多信息”2. 点击“仍要运行”3. 继续安装用 PowerShell 安装会直接下载官方版本，避开大多数浏览器下载限制。npm 全局安装

把它粘贴到 PowerShell 中。Linux

作者 Parsa Khazaeepoul，Pane 联合创始人。最后审查于 2026年10月.