Building Window
The Building Window opens when you click a building on the World Map and gives you everything you need to inspect or configure that building: its agents, its system prompt, its additional directories, its skills, and its MCP servers — all in one panel.
How It Works
Section titled “How It Works”Every building in a project — HQ, CodeForge, KnowledgeBase, WorktreeForge, Linear, MarketResearchers, GenericBuilding, and the specialized Council, LoopReview, and Frontend buildings — has the same outer shape: a window with a header, a content area, and a settings region. The header carries the building’s name (editable inline), its kind, and a close button. The content area is per-kind — what you see depends on the building’s purpose — but most building windows share a common set of sections you’ll recognize across kinds.
The window is draggable. Grab the header and drag it anywhere on the screen; it stays on top of the World Map and can be repositioned freely. Clicking a building elsewhere on the map closes the current window and opens the new one in its place.
Agent portraits
Section titled “Agent portraits”At the top of the content area, every building shows the agents stationed inside it as a row of agent portraits. Each portrait has the agent’s avatar, name, role, and a status dot. Click a portrait to open that agent’s Agent Chat window. The dot color tells you at a glance whether the agent is idle, thinking, streaming, or waiting on a permission or question.
Building prompt section
Section titled “Building prompt section”Most buildings have a system prompt that’s prepended to every agent inside them. The prompt section is editable: click into the text area, make changes, and save. Edits take effect on the next message any agent in the building sends.
Additional directories
Section titled “Additional directories”Buildings can have additional directories declared — folders outside the project’s root that agents in this building are allowed to read or write. This is how a CodeForge building can be pointed at a shared library or a Linear building at a doc folder. Add and remove directories from this section; agents inherit the building’s list on top of their own per-agent extras.
Skills picker
Section titled “Skills picker”The Skills picker lists every skill installed at the user level and lets you choose which ones are available to agents in this building. Skills picked here flow down to every agent unless an agent overrides the list in its own settings.
MCP servers section
Section titled “MCP servers section”The MCP servers section is one of three scopes Viberia supports — global (user-level), per-building (here), and per-agent. The list shows every MCP server currently attached to this building, with toggles for enable/disable, configuration buttons, and an “Add server” affordance that opens the MCP installation flow. See MCP Servers for the full model.
How to use
Section titled “How to use”A small How to use section near the bottom of most building windows gives a short, kind-specific summary of how to drive that building. For CodeForge it’s a quick reminder of the PRD → implement → review loop; for KnowledgeBase it’s the document workflow; for WorktreeForge it’s the worktree commands. This is documentation built into the app for the building’s intended workflow.
Webpage Mode
Section titled “Webpage Mode”Some buildings are configured for Webpage Mode — instead of (or in addition to) the standard content area, the window embeds an external URL inside an in-app webview. The built-in Linear building uses Webpage Mode to embed the Linear web UI alongside its three agents, so you can triage a backlog in the browser and chat with the Issue Writer at the same time. Custom buildings can be built in Webpage Mode from the Building Creation Wizard by picking the Webpage UI Mode at step six.
UI Indicators
Section titled “UI Indicators”- Status dots on each agent portrait — idle, thinking, streaming, or waiting.
- Editable name — clicking the building name in the header turns it into an input.
- Unsaved-changes dot appears in the prompt section when there are pending edits.
- Skill toggles highlight when a skill is enabled for this building.
- MCP status badges show running, stopped, or errored state per server.
- Webpage Mode banner appears at the top of the embedded webview to indicate that the content is external.
- Window drag handle — the header is grabbable; the cursor changes to a move icon.
- Close button on the right side of the header dismisses the window.
Use Cases
Section titled “Use Cases”- Open a specific agent. Click an agent portrait to jump straight into its chat.
- Adjust the team’s brief. Edit the building’s system prompt to change how every agent inside it behaves.
- Give the team more reach. Add an additional directory so agents can read or write outside the project root.
- Curate skills. Pick the slash commands that should be available to agents in this building, and exclude irrelevant ones.
- Add an MCP server. Plug the building into Linear, a database, or any other MCP server without affecting other buildings.
- Work side-by-side with an external tool. Use a Webpage Mode building (like Linear) to keep an external UI and an agent chat visible at once.
Limitations
Section titled “Limitations”- The agent portraits row is read-only from this window — you can’t add or remove agents here. Adding agents is done through the Building Creation Wizard when defining custom buildings; built-in kinds have fixed rosters.
- Webpage Mode requires the embedded URL to permit iframing. Sites that block embedding will fail to load.
- Skills picked at the building level can be overridden per agent. If a skill works for one agent but not another inside the same building, check the agent’s Skills tab.
- Editing the prompt while an agent is mid-reply doesn’t change the in-flight reply; it applies to the next turn.
- The building’s kind cannot be changed after creation. To switch kinds, destroy the building and place a new one.