autonomous_building

# Autonomous Application Building This skill teaches agents how to use PyWry MCP's agentic tools to autonomously design, build, and export complete widget applications from a plain-English description. ## When to Use These Tools Use the agentic tools when you need to: - Build a complete PyWry app from a description **without manual step-by-step tool calls** - Deliver a **full Python project** (main.py, requirements.txt, README.md, widgets/) - Interactively **gather requirements** from the user before generating code - Show **real-time progress** to the user during long build operations --- ## Available Agentic Tools ### `plan_widget` — AI-Planned Widget Spec Sends the description to the LLM via sampling and returns a validated `WidgetPlan` JSON object. Use this to **inspect the plan before committing** to building. ```python # Returns a WidgetPlan JSON (no widget is created yet) result = await client.call_tool("plan_widget", { "description": "A crypto price dashboard with symbol selector and refresh button" }) plan = json.loads(result[0].text) # plan contains: title, html_content, toolbars, callbacks, width, height, ... ``` ### `build_app` — End-to-End Autonomous Builder The primary tool for autonomous app building. One call: 1. Samples a `WidgetPlan` from the description 2. Registers the widget in the session 3. Returns `widget_id` **and complete runnable Python code** ```python result = await client.call_tool("build_app", { "description": "Task tracker w...