IWB Home Base

Inspiration

Small custom woodworking shops have to keep production moving while tracking estimates, installation dates, and potential work. The useful question is often simple: “What needs my attention, and what should I do next?” IWB Home Base explores how an owner and an AI agent can answer that question together inside one shared workspace.

What it does

The demo brings active projects, prioritized attention items, business opportunities, and follow-up drafts into one command center. The agent can inspect the business, investigate a specific project, and prepare a follow-up when the human asks. The resulting draft appears immediately in the same interface the owner is viewing.

For example, the agent can identify an unanswered estimate for Cedar House Bar, retrieve its project context, and prepare a follow-up for human review. Every draft is marked as requiring human approval. The demo has no sending capability.

Why WebMCP matters

Five structured browser tools turn the app into a collaborative workspace:

get_business_snapshot: understand the current business state.
get_attention_items: inspect priorities and recommended next steps.
get_project_context: investigate a specific project with bounded context.
get_active_opportunities: review potential work and follow-up opportunities.
prepare_follow_up: create a local draft requiring human approval.

The tools and the human interface use the same in-memory state. The agent can work with explicit project identifiers and structured context, and its draft becomes visible in the app without a reload. This reduces the need to copy project details into a separate conversation and manually transfer the result back.

How it was built

The standalone challenge edition uses React, TypeScript, and Vite, with deployment on Netlify. It registers five tools through browser WebMCP and connects them to the shared demo state. The repository includes source code, synthetic fixtures, local setup and validation commands, and an MIT license. The human interface also works in a standard browser; agent tool use requires a WebMCP-compatible environment.

Design challenges and lessons

The central design challenge was making agent action useful while keeping its scope clear. Read tools expose bounded business context, while the action tool only creates a local draft. The shared-state design makes the result visible to the human and keeps review part of the workflow. It demonstrates that useful human-agent collaboration depends on clear capabilities and visible outcomes as much as on natural-language interaction.

Impact and next steps

The intended audience is small project-based businesses whose owners juggle delivery, client follow-ups, and new work. This demo shows a concrete path from identifying an issue to preparing the next action. Future work could add persistent state, editing and approval workflows, and carefully scoped live integrations.

Challenge scope

I Wood Build Home Base existed before the challenge. This submission is a new standalone synthetic edition; the five-tool WebMCP integration, shared demo state, safety boundary, and challenge presentation are the challenge-specific work. All business data is fictional. There are no real customer records, live integrations, external APIs, or persistent storage.