✦ agent integration
Your agent. Every stack.
Give coding agents structured access to Better Fullstack. They can inspect options, validate combinations, preview file trees, and scaffold projects without guessing CLI flags.
✦ Install for
$ claude mcp add --transport stdio better-fullstack -- npx -y create-better-fullstack@latest mcprun in your terminal
Docsstructured tools
readable resources
configurable options
faster than prompt-only
✦ toolbox
Seven tools. One workflow.
The server exposes narrow, typed tools for guidance, schema lookup, compatibility checks, dry runs, creation, and adding features.
- 01
bfs_get_guidanceWorkflow rules, field semantics, and critical constraints - 02
bfs_get_schemaValid options for any category, filterable by ecosystem - 03
bfs_check_compatibilityValidate stack combinations with auto-adjustments - 04
bfs_plan_projectDry-run preview: generates the file tree in memory - 05
bfs_create_projectScaffold a new project to disk - 06
bfs_plan_additionValidate proposed addons for an existing project - 07
bfs_add_featureAdd features to an existing project
✦ readable resources
docs://compatibility-rulesWhich stack combinations are valid
docs://stack-optionsAll technology options per category
docs://getting-startedQuick-start recipes per ecosystem
✦ workflow
You describe. It builds.
Agents use the same schema and compatibility rules as the web builder, so generated commands match what users can configure visually.
- —frontend is an array: multiple frontends in one monorepo
- —"none" means skip, not "use the default"
- —Set ecosystem first: it decides which fields matter
- —Dependencies are never installed: your agent will tell you the install command
you: “Create a fullstack TypeScript app with Next.js, Hono, Drizzle, and PostgreSQL.”
→bfs_get_guidanceworkflow rules + field semantics
→bfs_get_schemavalid options for the stack
→bfs_check_compatibilitystack validated, 0 adjustments
→bfs_plan_projectdry-run preview, 59 files
→bfs_create_projectwritten to ./my-app
✓scaffold completerun bun install to finish
✦ benchmark-backed
2.6× faster than prompt-only.
In ScaffBench, MCP-guided project creation is faster and more reliable than asking an agent to hand-write a project from scratch.