Operations Guide
From purchase to first delivered work.
Everything you need to put a hired role or team to work in your own environment. No hosted runtime, no lock-in - you own the package and run it where you already run your AI.
1. Buy the role, team, or department
Pick a single role, a team that ships as one wired set, or a whole department. You own it at the version you bought, permanently. The Company Map recommends what to hire and in what order if you are not sure.
2. Install it into your own workspace
Each purchase is a downloadable package - the role files, sub-agents, references, and a generated install guide - that you drop into your own Claude Code or Claude project. Nothing runs on our infrastructure; it operates where you put it.
3. Map your tools
Every role names tool categories, never locked vendors. During setup you map each category to the tools you already run, so the role works against your real stack from the first task.
4. Set the approvals
Each role carries a capability contract: what it may do on its own, what it must run past you, and what it must escalate. You choose how cautious it is - ask before anything leaves the building, or move fast on low-risk work and review the rest. Privileged actions never happen automatically.
5. Put it to work
Give the role its first task with the starter prompts in the install guide. For a team, the Orchestrator routes work across the roles and the declared hand-offs pass context from one seat to the next, so the group delivers one outcome rather than fragments.
6. Keep it current (optional)
Updates are an optional subscription you add at checkout. Keep it active and you receive new versions, new sub-agents, and any new role added to the slice you own, with a per-client changelog. Stop it and you keep every version you already downloaded.
Q&A
Common questions about running the roles
Do the roles run on your servers?
No. A role is a package you install into your own AI workspace and run there. Your data, credentials, and approvals stay inside your own boundary - we never see your data.
Can a role take actions on its own?
Only within the approvals you set. Every role declares which actions are automatic, which need your sign-off, and which it must escalate. Privileged actions are never automatic, and each has a declared safe fallback.
What is the difference between a role and a team?
A role is one operating seat. A team is a group of related roles that ships with an Orchestrator and the hand-offs between them intact - it installs as one coherent set, not a folder of separate agents.
How do I know a role is any good?
Every authored role is independently audited against four pillars before it ships, and carries a version and a content hash so you can prove what you are running. See the audit and the trust model on the Security page.
Not sure where to start?
Build a free Company Map and we will recommend the roles to hire, in what order, on your stack.