Steward docs
How Steward works, in plain words. New here? Start with the owner guide. Its first section is all most owners need.
Start here
- Owner guideThe three things an owner does, and everything else when you want it.
- Getting startedJoin from your invite, the first messages to try, and how @steward and personal agents differ.
- Who does each setup stepWhat Steward asks a person for, and why. Its agents do the rest.
How it works
- ConceptsThe studio, businesses and charters, the board, cards, the ledger, budgets and trust.
- Agents@steward, personal agents, who can instruct whom, and how tasks become work.
- The ledger, Undo and moneyEverything that happened, undoing what an agent did, and each business's monthly numbers.
- LoopsOps, security and product work that repeats on its own, per business.
- Integrations and trustGitHub, Railway, email, social and Stripe, and how much @steward may do alone.
Running a business
- Launch to first dollarThe launch checklist, from domain and email to the first payment.
- Trying new ideasThe idea inbox, scoring, 14-day tests and verdicts.
- Service businessesClients, call notes, proposals, delivery and the retainer.
- The support loopThe inbox, inbound email, the contact form and escalations.
- The growth loopThe content calendar, email sequences, A/B tests and traffic by source.
- AnalyticsVisits, events and numbers, without cookies.
- Paid adsAd spend with a cap, and what each dollar brought.
Your own agent
- The connectorRun an agent on your own Mac, Linux or Windows computer, on your own Claude plan.
- Command lineThe studio in a terminal: cards, chats, tasks and agents, with --json for scripts.
- Bring your own modelClaude, GPT, Gemini, open and local models, per agent and per business.