How Brewyard works
A coordinated AI software team you brief in plain English, running on the Anthropic API key you already have. Three steps from sign-up to first useful output — and a live Mission Control to watch the work happen.
Three steps to a working team
Connect your Anthropic key
Paste your Anthropic API key once. We validate it with Anthropic, store the value encrypted in Azure Key Vault, and keep only a pointer in our database. The plaintext never lives in Postgres, never appears in a log, and never leaves the vault except to spawn your agents.
Create a workspace
Pick a name and set a monthly budget. You do not have to choose the team — the CEO reads your brief and hires what the work actually needs. If you want to change who is on it, you can see the line-up before it starts.
Brief your team
Tell the workspace what you want built or analysed. The agents discuss internally, delegate, and produce structured output — briefs, code, dashboards, decisions. You stay in the loop via the live dashboard and intervene whenever you want.
How work moves inside the team
The three steps above are what you do. This is what happens after you press send — and it happens without you in the loop, which is why the ceiling and the guard rails below matter.
You brief the CEO
One paragraph about your project, in your own words. No prompt engineering, no template, no format to learn. The CEO is the only agent you speak to by default.
The work gets split
A project manager turns the brief into workstreams and gives each one an owner. Which specialists exist depends on the brief: a competitive analysis hires a different team from a data import, and neither hires the whole catalogue. You can see the team it picked before it starts, and change it.
Specialists do their part
A business analyst gathers and reads. A tech lead judges feasibility. A CFO costs it against your ceiling. A QA engineer sends work back when it does not hold up. Each stays in its lane and says so when a question is not its own, so a question about money reaches the CFO rather than whoever happened to be holding it.
It comes back to you
Documents you can open, read and download. Every claim names the source it came from, and where nothing backed a claim the document says “not known” rather than filling the gap. The full trail of who did what, in what order, and what it cost sits behind the same screen.
Watch the team work in TeamForge Mission Control
Every Brewyard workspace opens into TeamForge Mission Control, the view of the system that runs your team: your org chart working in real time, tasks moving across a kanban board, and every message streamed to your screen as it happens.
- CEO
Read your brief. Splitting it into research, costing and a delivery plan.
- CEO
- CFO
- Project manager
- Tech lead
- Researcher
Ready to read
- —
Live org chart
See who is thinking, who is working and who is waiting — each agent lights up the moment it picks up a task.
A kanban of real tasks
The team breaks your brief into concrete tasks, and you watch them move from backlog to done as the agents deliver.
You approve the big calls
Anything consequential stops at an approval card first. Nothing irreversible happens without your explicit go-ahead.
Full transparency
The activity feed streams every agent message and tool call as it happens — nothing runs behind your back.
Under the hood
Each workspace gets its own private TeamForge runtime — an isolated container in Azure UK South (London) that scales to zero while you're away and talks to Anthropic with your key, fetched from the vault only at start-up.
The request path
Your browser
Mission Control dashboard, no installs
Brewyard control plane
Accounts, billing, orchestration
Your private runtime
TeamForge — an isolated container per workspace
Anthropic API
Billed on your key, zero markup
Key custody
- Your Anthropic key is stored encrypted in Azure Key Vault; our database only ever holds a pointer to it.
- The runtime reads the key at start-up via its managed identity — it is never passed around in plaintext.
- Log pipelines scrub anything that looks like an API key before it is written anywhere.
Your budget, guarded by the CFO
Set a monthly cap per workspace. The CFO agent tracks spend in real time, warns you at 75% and pauses the team at 95% — an end-of-month surprise is structurally impossible.
Runtime hours, counted
Your plan includes a set number of hours each month, counted across your account rather than per person. You can see what is left before you start, not after the invoice.
What you can rely on
Your key, your control.
Brewyard never charges you a markup on Anthropic usage. The CFO agent warns at 75% of your monthly cap and pauses agents at 95% so you never get an end-of-month surprise.
Your data, your retention.
Workspace content lives in Azure UK South (London). Cancellation deletes personal data within 30 days, with the legal-retention exceptions noted in the Privacy Policy.
Transparent by design.
You cannot read our source, and we would rather say so than imply otherwise. What you can do is watch: every agent message and every tool call is streamed to your screen as it happens, and every run leaves an audit trail you can export. Our data flow and sub-processor list are published on this site rather than offered on request.
Grounded, or it says so.
Agents have to be able to point at where a claim came from, and where nothing supports an answer the answer is that nothing supports it. This is a rule of the runtime rather than a line in a system prompt you have to hope was followed — a confident guess is the failure mode that costs you the work, not a rough edge.
Or try it on a project first: a one-off day pass — no subscription to cancel, and auto-renew is a switch you control.