How Rudder Works

You do not need to learn the whole product before using Rudder. Start with five ideas.

​

1. An organization is your team’s workspace

It contains the Agents, conversations, tasks, projects, files, and settings for one team or area of work. If you use Rudder alone, create one organization for your personal work and keep going.

​

2. An Agent is a reusable worker

An Agent has a role and a runtime. The role says what kind of work it should handle. The runtime is the program that does the work, such as Codex, Claude Code, a local command, or a webhook. Each time the Agent works, Rudder creates an Agent Run. The Agent stays; a Run is only one attempt. Open the Run when you need to see progress, usage, errors, or lower-level details.

​

3. Work can start in Chat or an Issue

Use Chat when you want to ask, clarify, and revise in one conversation. Use an Issue when the task needs a clear owner, status, dependency, or reviewer. An Issue is useful structure, not a requirement for every task.

​

4. The result belongs with the request

An Agent may return an answer, change files, create a report, capture a screenshot, or provide a link. Rudder keeps these results connected to the Chat or Issue that asked for them, so you can understand them later.

​

5. People still decide what happens next

You can ask for another revision, accept the result, assign a reviewer, mark a task blocked, or create follow-up work. Rudder makes those decisions visible; it does not silently decide that every Agent response is correct.

​

Add structure only when it helps

Dashboard and Calendar help you scan what is happening. They point back to the original Chat, Issue, Run, or Automation rather than creating a second copy of the work.

​

What to read next

1. Run your first task.
2. Choose Chat or an Issue.
3. Understand Agents and Runs.
4. Use an Issue from start to finish.
5. Review Agent work.