deploy: PR #42 preview

Author: github-actions[bot]

pr/42/about.md

@@ -0,0 +1,63 @@
+## Model Context Protocol Server
+Configure experimental MCP server under Blockbench settings: __Settings__ > __General__ > __MCP Server Port__ and __MCP Server Endpoint__
+
+The following examples use the default values of `:3000/mcp`
+
+> __Be sure your Blockbench settings match the port and endpoint used in the MCP JSON configuration.__
+
+## Installation
+
+### Claude Desktop
+
+__`claude_desktop_config.json`__
+
+```json
+{
+  "mcpServers": {
+    "blockbench": {
+      "command": "npx",
+      "args": [
+        "mcp-remote",
+        "http://localhost:3000/mcp"
+      ]
+    }
+  }
+}
+```
+
+### VS Code
+
+<span style="background: darkred; box-shadow: 1px 1px 6px rgba(0 0 0 / 50%); color: pink; font: 600 12px sans-serif; padding: 0.3em 0.5em; border-radius: 0.25em; outline: 0.125em solid red;">Unstable</span>
+
+> This plugin _should_ be compatible but Copilot often throws errors when using most tools.
+
+__`.vscode/mcp.json`__
+
+```json
+{
+    "servers": {
+        "blockbench": {
+            "url": "http://localhost:3000/mcp"
+        },
+    }
+}
+```
+
+## Basic Usage
+
+1. Start Blockbench
+2. Start your MCP client
+3. Enable/disable tools in client
+4. Prompt to create models
+
+### Limitations / Expectations
+
+Only a small fraction of Blockbench's API has been mapped to a MCP server tool, so some capabilities may be limited.
+
+"One-shotting" a model is possible but human-in-the-loop interaction and project clean-up should usually be expected.
+
+### Tips
+
+- __Set the agent up for success__ by adding textures and groups to the project ahead of prompting.
+- __Find the right amount of context__ by disabling tools you do not want or need, _but also_ try providing lots of specific project details and even __use other MCP servers in the same context.__ (i.e. File or web access tools.)
+- __Make a backup.__ This is experimental tech! Working in existing projects may yield unexpected results.