AI Agents

Better Fullstack supports three AI workflows:

• Generate project-local instruction files with --ai-docs.
• Install the Better Fullstack agent skill so coding agents use the CLI instead of hand-writing starter files.
• Let an AI coding agent use the Better Fullstack MCP server for schema lookup, compatibility checks, dry-runs, and project creation.

Preferred workflow

1. Choose the target ecosystem: typescript, react-native, rust, python, go, java, or elixir.
2. Use the Better Fullstack agent skill, MCP server, or explicit CLI reference.
3. Validate compatibility before scaffolding.
4. Dry-run to inspect the file tree.
5. Scaffold with dependency installation disabled when an agent is driving the flow.
6. Tell the user the exact install, test, and run commands.

Generated AI docs

Use --ai-docs to generate project-local context files for coding agents:

Multiple values are accepted:

npm create better-fullstack@latest my-app -- --ai-docs claude-md agents-md

Use --ai-docs none to disable these files.

CLI workflow

For agents without MCP access, pass explicit flags and avoid side effects:

npm create better-fullstack@latest my-app -- \  --ecosystem typescript \  --frontend next \  --backend self \  --runtime none \  --database postgres \  --orm drizzle \  --auth better-auth \  --api trpc \  --db-setup none \  --ai-docs agents-md \  --version-channel stable \  --no-install \  --no-git

Agent skill workflow

The Better Fullstack agent skill lives in this repository and teaches supported agents to use the CLI with the benchmarked fast workflow: map the stack, run a dry-run, scaffold without install or Git side effects, then report follow-up commands.

See the agent skill setup guide for Codex, Claude Code, Cursor, GitHub Copilot, and Gemini CLI install commands.

AI provider options

TypeScript projects can select one AI option with --ai: vercel-ai, mastra, voltagent, langgraph, openai-agents, google-adk, modelfusion, langchain, llamaindex, tanstack-ai, ai-cli, or none.

Python projects use the separate multi-select --python-ai option with langchain, llamaindex, openai-sdk, anthropic-sdk, langgraph, crewai, or none.

Dependency placement depends on the selected framework and backend. Server-side packages are added to the server package or self-backend app; supported frontend integrations receive their matching client hooks or packages.

MCP workflow

The Better Fullstack MCP server exposes schema lookup, compatibility checks, dry-run previews, project creation, and feature additions as structured tools. See the MCP setup guide.

This is distinct from the generated-project mcp addon. The server helps an agent create Better Fullstack projects; the addon writes MCP configuration inside a generated project.