Frequently asked questions
The questions people actually ask in their first week.
Using it
What do I actually type? Describe an outcome, not a task. "Get Acme moving again — they went quiet after the pricing call" works better than "write an email to Acme", because it tells your agents what you are trying to achieve and lets them notice when a step is wrong.
Does it do things without asking me? Yes — that is the point — and you decide how far that goes. Anything that leaves your workspace, spends money, or touches production pauses for a one-tap approval until you pre-authorize it with a standing grant. Everything else just gets done and shows up on your Dashboard.
What is an objective, versus just chatting? Chatting gets you an answer. An objective is work your team runs: it gets decomposed into tasks, delegated to department chiefs, and produces deliverables. Your Chief of Staff proposes one when your request is bigger than a reply.
Can I change my mind halfway? Yes. Steer a running response with a correction, pause an objective, retry a failed task, or stop it entirely. You are never locked in.
Why did it ask me a question instead of just doing it? Because the answer would have changed the work. Agents are told to ask rather than invent — an invented detail in your first deliverable is worse than a clarifying question.
Money
What is free? Chatting with your Chief of Staff. Scoping and running objectives needs an active plan.
What is the difference between a plan and AI credit? A plan is your licence to run work — it sets seats, concurrency and included channels. AI credit pays for tokens. You need a plan either way; credit only applies if Foreman is funding the AI rather than your own provider keys.
If I bring my own keys, do I still need a plan? Yes. Your keys change who pays for tokens, not whether you can run work.
What happens if my balance runs out mid-objective? Work pauses. It does not fail and it is not lost. Top up and it resumes. Auto top-up exists so a long run does not stall overnight.
Which tier should I pick? Size by concurrency and seats, not credit. Credit is easy to top up; concurrency is what decides whether five things run at once or queue behind each other.
Data and control
Where does my data live? In your workspace, isolated from every other workspace. You can also connect your own database and run the workspace on it, or export everything.
Do you train on my data? No.
Can I get my data out? Yes — a full workspace export, plus CSV export from the CRM.
Who on my team can see what? Three roles and about forty individual capabilities. Members deliberately do not get billing, financial data, CRM schema changes, workspace memory or database configuration. You can grant individual capabilities without promoting anyone.
Integrations and channels
What can it connect to? Hundreds of tools across nine categories, plus email, team chat and social messaging. Foreman also ships its own CRM, so you do not need an external one.
Can agents write to my tools, or only read? Read by default. Writing is a separate, explicit grant per action — a wall, not a request that agents behave.
Does it post to LinkedIn for me? No, and that is deliberate. Foreman drafts and queues LinkedIn touches; you paste and send. It never operates your account.
Studio
Where does the code run? On your own host — your desktop app, or a cloud host Foreman provisions. Not in a sandbox you cannot reach.
Can it deploy? Yes, to Render or Vercel, and it can push to GitHub, GitLab or Bitbucket.
Is it a real codebase? Yes. Files on a real machine, in a real repo, that you can open in your own editor.