herdr orchestrator, 여러 에이전트를 슬래시 커맨드로 돌리는 법
Based only on the official documentation checked as of 2026-09-03, the practical takeaway is an automation and CLI workflow that lets a script or one agent create work for other coding agents, inspect their state, and collect results. This article explains that documented route first, then records the narrow personal workflow I use.
What does the orchestration angle mean?
Short answer: “herdr orchestrator” is not a separately documented product name; it is the search angle used here for the documented automation layer and CLI coordination flow for coding agents.
The official Agent automation page describes a script or one agent creating work for other agents, checking their state, and collecting their results. Its documented responsibilities are divided as follows.
- Layout controls workspace, tab, and pane topology.
- Pane controls the raw terminal.
- Agent controls the lifecycle of a recognized coding agent.
Starting an agent requires an existing shell pane; starting the agent does not create or rearrange the layout. The automation CLI documents lifecycle states including working, blocked, done, idle, and unknown.
When the documentation specifies creation-command output, a script should capture the returned JSON IDs rather than predict them. Agents stay in real terminal panes and their state is collected in the sidebar, so the official automation guide is the route for coordination from a script or another agent.
How does the official skill/CLI flow relate to slash commands?
Short answer: The verified official pages do not provide a literal /... command for invocation.
The documented route is a reusable agent skill plus the local CLI. A coding agent running in a Herdr-managed pane with HERDR_ENV=1 uses the herdr CLI to control the runtime. Unless a primary agent document establishes otherwise, a slash-command menu in a personal workflow belongs to the coding agent’s skill or custom-instruction UI, not to a Herdr builtin.
The reusable skill is installed with this official command.
npx skills add herdrdev/herdr --skill herdr -gOmitting -g makes the installation project-local. The skill lists inspection of workspaces, panes, and neighboring agents; pane splitting; output reading; waiting; and helper starts. The CLI uses the same local socket API as the running server uses for integrations and agents, and most commands return JSON for deterministic scripting.
Plugins are executable workflow packages whose manifests can declare actions and which the runtime can launch. The full CLI is their plugin API, but that does not establish any literal slash-command syntax.
How do you get started?
Short answer: From a project directory, run the runtime command to launch or attach to the default background session.
herdrIf the session has no workspaces, one opens automatically. A workspace is the project-level container for tabs, panes, and agents. Start a supported coding agent in a pane and the runtime detects it automatically; the sidebar shows working, blocked, done, or idle across workspaces.
Agents keep running after the client detaches. The install page lists direct installers for Linux/macOS and Windows, along with Homebrew, mise, and Nix routes; those options are secondary here because the workflow starts with the project-directory command.
How is the personal use limited?
- I currently subscribe to AGY, Cursor agent, and Codex CLI.
- I run them through this runtime.
- I created slash commands and use that workflow.
What should the FAQ answer?
Short answer: The four checks are whether the name is official, whether an invocation method is documented, whether automation can coordinate agents, and what happens after the client detaches.
| Question | Answer |
|---|---|
| Is “herdr orchestrator” an official product name? | The exact name was not found on the verified official pages. Here it is the search angle for the documented automation and CLI workflow. |
Does the official documentation give a /... command to invoke? | Not on the verified official pages. Use the documented herdr CLI and the agent-skill installation facts as the reference point. |
| Can a script coordinate agents? | Yes. The official Agent automation page explicitly presents that use case and separates layout, pane, and agent controls. |
| Do agents keep running after the client detaches? | The official Quick start and Concepts pages say that the server and agents continue after client detachment; stopping the server ends its panes. |
What are the sources?
Short answer: The article uses only official documentation checked as of 2026-09-03.
- Official Agent automation documentation — describes the agent automation layer and coordination flow.
- Official Agent skill file documentation — documents reusable-skill installation and CLI use from a managed pane.
- Official CLI reference — documents the CLI and JSON responses.
- Official Quick start documentation — documents starting a session from a project directory.
- Official Agents documentation — documents terminal panes and sidebar state.
- Official Concepts documentation — documents server and session concepts.
- Official Plugins documentation — documents executable workflow packages and the CLI API.
- Official Install documentation — lists the supported installation routes.
- Herdr documentation — provides the official install, learn, and configure guidance.
If tool subscriptions feel expensive, see only the discount route.