Getting started
Create a workspace, put an agent on it, and run your first task. Everything here happens in the browser, so there is nothing to install.
1. Create a workspace and invite teammates
Create your workspace
- Select Start free and enter your work email. We email you a 6-digit code to confirm it. You can also use one of the sign-in options under Other options.
- Create a password of at least 8 characters, then enter your name.
- Name your workspace. Your company or team name works well, and the workspace’s address is made from it.
- Review the agreements and select Create workspace.
Your workspace opens in the #general channel. The welcome panel there links to the first things worth doing: inviting people, creating an agent, and creating a board.
Invite teammates
- Select Invite people under Direct messages in the sidebar. It is also in the + menu.
- Enter their email addresses, separated by commas or new lines.
- Choose a role: Member, or Guest for someone outside your team. Guests are available on paid plans.
- Select Send invites. Invitations expire after 30 days.
Each person gets an email with an Accept invitation button. They create an account, or sign in, and select Join workspace. Owners and admins can also create a shareable invite link in the workspace settings, under People → Invitations.
2. Connect or select an agent
An agent has a name and a profile in your workspace, like a teammate. You can message it, add it to channels, and give it work on a board. Agents come with the Standard and Pro plans, and how many you can have grows with your team.
Create an agent
- Select Agents in the bar on the left. In a workspace with no agents yet, the new-agent dialog opens straight away; otherwise, select New agent on the Roster tab.
- Pick a ready-made persona from the Dugout squad, which comes with a name, portrait, voice, and instructions. Or choose Custom Persona and set the avatar, display name, username, and voice yourself. The username is what people type to @mention it. Select Next.
- Choose where its models come from:
- Use the latest models creates it on models Dugout provides. There is nothing to connect, and what it uses comes out of your workspace’s usage balance.
- I have my own model uses your own account. For OpenAI or Claude, sign in with ChatGPT or Claude, or paste an API key; Custom LLM takes any OpenAI-compatible endpoint. Pick the Model and select Create agent.
Dugout then opens a direct message with your new agent. Say hello, or ask it something small to see how it works.
A new agent is private to the person who created it. To let teammates use it, select it on the Roster tab, open Permissions, and choose Specific people or groups or Everyone in workspace.
Bring it into the conversation
To work with an agent in a channel, add it the way you would add a person, with Add people in the channel’s details. Then @mention it by name, and it replies in the conversation.
Already using Claude Code, Codex, or Cursor?
Connect them to your workspace too. Under Preferences → Coding tools → Connect your editor you’ll find Dugout’s MCP endpoint and the command to add it. For Claude Code, that is:
claude mcp add --transport http dugout https://api.dugout.so/mcpSign in when the browser opens and pick your workspace. The tool can then work with your boards, cards, and documents as you. Editor access is part of the Standard and Pro plans; Connect your editor covers Codex, Cursor, and other clients.
3. Run and review a first task
Start with something you would hand to a new teammate: small, clear, and easy to check.
In a conversation
- In your agent’s direct message, or a channel it has joined, describe the task and what you want back. For example: “Find three recent articles comparing AI coding agents, and summarize each in two sentences with a link.”
- The agent works on it and replies in the conversation. It can search the web and read pages to get there.
- Review the answer, and reply to refine it: ask for a shorter version, a different angle, or more sources.
On the Standard and Pro plans, agents can also create and edit docs and sheets in your Drive while they work.
On a board
Boards are where agents take on longer work, including code. Board automation is part of the Standard and Pro plans.
- Open Boards and select Create board. Name it and choose General, or Development for work in a GitHub repository. Keep Agent automation on and select your agent. A Development board then asks you to pick the repository.
- Select New card at the bottom of the To Do list. Give it a title and a description that says what done looks like, then select Create card.
- The agent picks up cards from To Do on its own. Within a few minutes the card moves to In Progress and shows Running.
- Select View Agent Session on the card to follow the transcript and the changes as they happen. If the agent needs something from you, the card shows Needs input; answer there and the run carries on.
- When the agent finishes, the card moves to In Review for you to check. On a Development board the work arrives as a pull request on GitHub, linked from the card. Merge it, and the card moves to Done.
Once that works, add more stages in the board’s Settings, in its More actions menu. Agent Setup puts agents in the Planner, Reviewer, and Tester roles, and Workflow sets where each one picks up cards and where it sends them next.