Documentation

How Harness works in practice — routines, capabilities, approvals, connectors, licensing, and what to do when something does not behave. Written so you can run it without asking us.

Start here

First steps

Getting started

Install, activate, connect your AI provider and get the first routine finished end to end.

Read

Routines & scheduling

Write a routine in plain language, test it once, then put it on a schedule that runs without you.

Read

Capabilities & permissions

The default-deny model: what a capability is, how to grant one, and how to keep the blast radius small.

Read

Control & connections

Keeping the work inside the rules

Approvals & audit

How the approval queue works, what pauses in it, and how to read the record afterwards.

Read

AI providers & keys

Connect your own Claude or GPT account or key, and see exactly what is sent where.

Read

Files, email & systems

Point Harness at folders, mailboxes, databases and APIs — one connection at a time.

Read

Account & problems

Licence, updates and fixes

Licence & activation

One licence per PC: how activation works, moving to another machine, and what happens if updates lapse.

Read

Troubleshooting

A routine stopped, a source was unavailable, a step failed — how to see why and get it running again.

Read

Prefer to watch

The same steps, on screen

Short walkthroughs cover setup, building the first routine and working the approval queue.