Why does Pane use a CLI instead of only MCP?

Every coding agent can run shell commands, so a CLI gives Pane an agent-agnostic control surface without requiring per-agent MCP setup. It is token-efficient, self-discoverable, easy to script, and works across Windows, WSL, macOS, Linux, and remote hosts.

Does Pane support MCP?

Pane works with agents that use MCP for GitHub, Linear, Jira, Slack, and other outside tools. Pane itself exposes local workspace control through runpane because the terminal is the common layer every CLI agent already has.

Should I install runpane globally?

For a one-time install or update, use npx --yes runpane@latest, pnpm dlx runpane@latest, or pipx run runpane. Use a persistent global install when you want the command available repeatedly for agents or scripts.

Can agents use runpane from WSL when Pane is installed on Windows?

Yes. If Pane is running as a Windows app and the agent is inside WSL, call the Windows wrapper through powershell.exe and use a Windows working directory such as $env:TEMP before running runpane.

How does runpane enable agent loops?

Agents can discover the command contract, register repositories, create panes, add reviewer tabs, wait for terminal readiness, read compact screen state, submit text or composers, and validate progress before reporting success.

<- back to runpane.com

runpane cli

the agent-operable cli for pane

Pane is the only agent manager built around true cross-OS support: Windows, WSL, macOS, Linux, local machines, remote hosts, and any CLI agent that runs in a terminal. runpane is the small command surface that lets humans install Pane and lets agents operate Pane.

true cross-os support

Windows and WSL are first-class, not an afterthought. The same workflow also runs on macOS, Linux, VMs, and remote hosts.

npm, pnpm, and pypi

Start with npx, pnpm dlx, pipx, or a persistent install when agents need a stable command on PATH.

agent-operable cli

Agents can discover Pane's schema, register repos, create panes, inspect terminal output, and send input.

loop-ready primitives

Existing agent CLIs can use Pane as the visible control plane for loops without waiting for a hosted scheduler or model-specific SDK.

install posture

For humans installing or updating Pane, the recommended path is a one-time runner. It always resolves the latest package and avoids leaving global state behind:

npx --yes runpane@latest
pnpm dlx runpane@latest
pipx run runpane

For agents, scripts, or repeated local automation, a persistent install is reasonable because the agent can call runpane directly:

npm i -g runpane
runpane setup

python -m pip install runpane
python -m runpane setup

why cli over mcp for pane control

MCP is good for external tools. Agents can use MCP for GitHub, Linear, Jira, Slack, docs, search, and other systems. Pane does not need to replace those integrations.

Pane control is local and terminal-native. Every coding agent already knows how to run shell commands. A CLI works in Codex, Claude Code, Aider, Goose, OpenCode, scripts, cron, and CI without a separate server install.

The contract is token-efficient. runpane agent-context prints a brief schema by default. Agents can lazy-load one detailed command only when they need it.

The contract is public before install. Agents can read /runpane-cli-contract.json or the docs reference before running any package manager command.

The workflow is packaged as a skill. The open-source Codex and Claude Code `runpane-orchestrator` skills capture the full issue-to-PR loop.

The same command works for humans and agents. The installer, updater, diagnostics path, repo registration, pane creation, terminal output, and terminal input all live under one inspectable surface.

the core loop commands

runpane agent-context
runpane doctor --json
runpane agent-context --command "panels wait" --json

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

runpane panes create --repo active --name "issue-252" --agent codex --prompt "Plan this issue" --source agent --wait-ready --yes --json
runpane panes list --repo active --json
runpane panels list --pane <pane-id> --json
runpane panels wait --panel <panel-id> --for ready --json
runpane panels screen --panel <panel-id> --limit 80 --json
runpane panels output --panel <panel-id> --limit 200 --json
runpane panels submit --panel <panel-id> --text "Continue" --yes --json
runpane panels submit-composer --panel <panel-id> --strategy auto --yes --json

wsl and windows pane

If an agent runs inside WSL while Pane is installed on Windows, call the Windows wrapper through PowerShell and avoid UNC current-directory issues by setting a Windows working directory first:

powershell.exe -NoProfile -Command \
  'Set-Location $env:TEMP; runpane repos list --json'

where to go next

- Set up your first AI agent loop

- Read the runpane CLI reference

- Fetch the machine-readable CLI contract

- Use the RunPane Orchestrator skill for Codex

- Use the RunPane Orchestrator skill for Claude Code

- Use the compact agent loop docs

frequently asked questions

Download*

Windows SmartScreen warningDirect downloads can show a SmartScreen warning while Pane is unsigned. Pane is fully open source, so you can audit the code and build from source yourself.1. Click More info2. Click Run anyway3. Continue the installerThe PowerShell install downloads the official release directly and avoids most browser download friction.npm global install

Paste that in PowerShell.MacLinux