Begins with one developer. Indispensable for a team.
A solo developer can watch one agent. A team runs many — across many products — and that's where orchestration and observability stop being a nicety and become the job. Fragua is built so more than one person can share a workspace without stepping on each other.
Many people, many products, one workspace each.
The hard part of multi-agent work isn't running the agents — it's two people touching the same repository at the same time. Fragua serializes the dangerous parts and isolates the rest.
Roles on the model, features in parallel.
Authorization lives on the records as plain predicates — admin or member, who can edit, delete, or drive a phase — so there are no scattered permission checks to drift out of sync.
Each feature executes in its own git worktree on its own branch, so different features run side by side. Every spec carries who created it and the branch it produced, so you can see the whole team's parallel work — and two people never fight over the same draft.
Amenities selection for properties
ImplementingSelect amenities on new and edit, from the existing catalog only — and a property now needs at least one to publish.
Danger zone to Properties
ShippedA danger zone at the bottom of the editor — unpublish a live property, or delete it and its images when it isn't published.
Preview properties
DraftA user editing a property needs to see a preview of how it will appear on the public page.
Owner property listing
ShippedOwner goes from blank to a complete property page on a phone in one sitting — profile, photos, amenities, pricing, availability.
The budget owner sees the whole account.
Spend rolls up per workspace, per account, and per month. Execution runs are bounded per workspace, so a burst of starts queues cleanly instead of stampeding the host.
The roll-up is the same artifact a solo developer sees, widened to the team — every workspace on one card, each with its run count and the shape of its spend, not one person's runs in isolation.
By workspace
16 runs-
Almighty monolithlast: $0.39$1.427 runs
-
Marigoldlast: $0.31$1.185 runs
-
Nimbuslast: $0.22$0.943 runs
-
Cartographerlast: $0.13$0.711 run
Context lives in the system, not in someone's head.
Every workspace accumulates a shared memory: the Knowledge Base you upload, the durable artifacts each phase produces, and the complete history of every run. When someone onboards, hands off, or leaves, the context the agents work from doesn't leave with them.
That's the difference between a team's tooling and a personal one. The plan, the references, the decisions, and the receipts are all in the workspace — readable by the next person and the next agent alike.
Documents
5-
Mock Property Edit
62.1 KB · by Daniela Ruíz · 9 days ago -
Mock Property listing
25.8 KB · by Daniela Ruíz · 9 days ago -
halua-icon
624 Bytes · by Daniela Ruíz · 12 days ago -
halua-horizontal
6.73 KB · by Daniela Ruíz · 12 days ago -
Mock Sign In
14.6 KB · by Daniela Ruíz · 12 days ago
Feature specs
Finer-grained team controls are coming.
A read-only viewer role and per-person cost and run attribution — who triggered which run, and what their share of the spend was — are on the way. We'd rather ship them honestly than claim them early. Today, roles are member and admin, and cost is attributed to the workspace and account.
Bring the team.
Fragua is in private beta. Tell us about the team and the products you're building.