How Harness, Switchboard, and Cairn fit together

Harness runs your agents, Switchboard hands them work, and Cairn holds what they make. Each is useful alone; together they close a loop from an event to a result a person can read. The full walkthrough lives on the Harness docs, so this page only says what each piece is and is not.

	Is	Is not
Switchboard todo	A dispatch lease: work from a verified event that one agent claims under a lease and carries to complete or fail	A second tracker: the issue or pull request stays the record of the work, and the todo points an agent at it
Cairn artifact	Evidence: the review, diff, log or trace an agent made, at a stable link you cite from your tracker or a todo's result	Where work is tracked or where the next step is decided
Harness	A supervisor: it starts, restarts, schedules and attaches to agent processes, and keeps their logs and run history	A task manager: it does not decide where work comes from or where results go

Read the canonical page: How Harness, Switchboard and Cairn fit together on the Harness docs walks one loop end to end and links the setup guide for each step.

Setting up the Switchboard side starts at Concepts, then your first endpoint.