Model Context Protocol

MCP Server

Let AI agents scaffold and modify fullstack projects programmatically. Works with Claude, Cursor, VS Code, and any MCP-compatible client.

Setup

Register the server with your AI agent. No hosting required — runs locally via stdio.

Terminal
claude mcp add --transport stdio better-fullstack -- npx create-better-fullstack mcp

7 Tools

Discover, validate, preview, and create — all via structured tool calls.

bfs_get_guidance

Workflow rules, field semantics, and critical constraints

bfs_get_schema

Valid options for any category, filterable by ecosystem

bfs_check_compatibility

Validate stack combinations with auto-adjustments

bfs_plan_project

Dry-run preview — generates file tree in memory

bfs_create_project

Scaffold a new project to disk

bfs_plan_addition

Validate proposed addons for an existing project

bfs_add_feature

Add features to an existing project

3 Resources

Read-only context agents can fetch for reference.

docs://compatibility-rulesWhich stack combinations are valid
docs://stack-optionsAll technology options per category
docs://getting-startedQuick-start recipes per ecosystem

How it works

Just describe what you want. The agent calls the right tools in order.

You say:

“Create a fullstack TypeScript app with Next.js, Hono, Drizzle, and PostgreSQL”

1
bfs_get_guidance

Learn the rules

2
bfs_get_schema

Discover valid options

3
bfs_check_compatibility

Validate the stack

4
bfs_plan_project

Preview file tree

5
bfs_create_project

Scaffold to disk

Tips

  • frontend is an array — supports multiple frontends in one monorepo
  • "none" means skip, not “use the default”
  • Set ecosystem first — it determines which fields are relevant
  • Dependencies are never installed — the agent tells the user to run install manually