Use GitHub Copilot CLI agents

Your coordinator can start agents that run GitHub Copilot CLI. Your coordinator is the agent you talk to, which starts the other agents, hands out work and reports back to you. They send and receive messages the same way your Claude Code agents do, so the coordinator hands them work and reads their replies.

Before you start

• Thrum set up in your repository, with a running coordinator. The quickstart starts one on Claude Code. That is the recommended way to begin, and the steps on this page assume it.
• Install the Copilot CLI on the machine where the agent will run: npm i -g @github/copilot.
• Sign in to Copilot the way you do today. Thrum does not sign in for you.

What it looks like

This is an example conversation. Your coordinator’s wording will differ.

You:
  Start a GitHub Copilot agent to write the regression tests for the login handler.

Coordinator:
  Created impl_login_tests on GitHub Copilot CLI, in its own worktree. It has read its briefing and has the task. I'll tell you when it reports back.

You:
  Which tool is impl_login_tests running?

Coordinator:
  GitHub Copilot CLI. I run Claude Code.

You do not message the Copilot agent yourself. The coordinator does.

Add GitHub Copilot agents through your coordinator

1. Ask your coordinator (in its terminal, or in My Inbox, your inbox in Thrum’s web UI):

   Install the Thrum plugin for GitHub Copilot.

   Do this once per machine.

   Each time Thrum starts a Copilot agent, it sets up the small part of the plugin that points the agent at its saved notes. The plugin you are asking for adds the rest: the Thrum commands and role playbooks (skills), including /thrum:prime, which loads the agent’s briefing. The briefing is the summary an agent reads at the start of a session.

   Signing in to Copilot stays with you.

2. Ask your coordinator to create the agent, as in the example. The New agent dialog in the web UI has no tool choice and always starts Claude Code, so for Copilot you ask the coordinator in plain words.

3. Give the coordinator the task. It sends the task to the Copilot agent and brings you the result.

4. Check it worked. Open the Agents page and select the agent, then open its Terminal tab. You see Copilot’s own screen. Thrum types /thrum:prime for the agent, and its reply says it is primed. Or ask the coordinator: Is the Thrum plugin installed in GitHub Copilot? or Which tool is each agent running?

If your coordinator will run in GitHub Copilot

Skip this section if your coordinator runs in Claude Code. If GitHub Copilot will be your first agent, there is no coordinator to ask yet, so install the plugin yourself. thrum-team/thrum-pro is where the plugin is distributed from. It holds the plugin’s text files, not the Thrum program. Free and Pro both use this same plugin. Thrum is published by THRUM LLC and built by Leon Letto (About). The first command adds that repository to Copilot as a plugin source named thrum. The second installs the plugin thrum from it. Run these once per machine:

copilot plugin marketplace add thrum-team/thrum-pro
copilot plugin install thrum@thrum

Run copilot plugin list. It must show thrum.

To update the plugin, run these:

copilot plugin marketplace update thrum
copilot plugin update thrum

What Thrum does for you

When the coordinator starts a Copilot agent, Thrum:

• Runs copilot in a tmux session, in the agent’s own worktree. A worktree is a separate checkout of your repository on its own branch. tmux keeps each agent running in its own terminal session.
• Passes the agent’s role, and the model and effort if you named them.
• Types /thrum:prime into the session so the agent loads its briefing. The Thrum plugin provides this command. Without the plugin, Copilot answers “Unknown command”.
• Answers Copilot’s “trust this folder” question for the agent.
• Starts Copilot with Thrum’s commands, common read commands and common git commands already allowed, so routine work does not stop to ask.

The agent uses Thrum by running thrum commands in its shell. Thrum writes a short instruction file named AGENTS.md into the agent’s folder. It lists the Thrum commands the agent can run.

Tips

Tip

Name the model in the same message when it matters: Start a Copilot agent on <model name>. Replace <model name> with the model you want. If you leave the model out, Copilot picks its own default.

Note

Copilot can still stop for a permission prompt that Thrum does not pre-allow. Answer permission prompts shows how the coordinator handles these.

Note

Thrum keeps its Copilot support files up to date each time it starts an agent. To update the plugin itself, ask your coordinator: Update the Thrum plugin for GitHub Copilot.

Next steps

• Mix runtimes and models: put each task on the tool and model that fits it.
• Run agents in tmux: where the agent’s session lives and how to watch it.
• Messaging between agents: how the coordinator and the Copilot agent talk.