Skip to main content
New to this? Follow the step-by-step setup, or let a coding agent do it for you.

​
Create an app

Scaffold a new FrogBot project with everything wired up:
The setup asks you to choose a database, an AI provider, the provider’s API key (or whether to use the one already set in your environment), and any coding agents that should receive the FrogBot skill. The blank template is selected automatically while it is the only available template. Your current package manager is detected. Dependencies are installed first, then a Git repository is created with an initial commit that includes the lockfile, so the project is ready to run when the command finishes. Pass --no-install to skip installing dependencies, or --no-git to skip Git. If the install fails, the command stops before Git and exits with an error. See Coding agents for skill installation options, agent selection flags, and project paths. For a non-interactive setup with the defaults, use --yes:
The defaults create a Next.js app using SQLite with no AI provider, agents, or chat. Choose a provider with --ai zen, --ai openai, --ai anthropic, --ai google, or --ai bedrock. Pass the key with --api-key to write it to .env. Without prompts, a key already set in your environment is not copied into the app; the scaffolder tells you it found the key and did not copy it. Select another database with --db sqlite, --db postgres, or --db mongodb.

​
Command options

Any setup choice you pass as a flag is not asked. The command also skips every prompt when --yes is set or when it doesn’t run in an interactive terminal, and uses the defaults below for anything you didn’t pass. An unknown option or a missing value exits with an error. An invalid choice exits with an error that lists the valid values.

​
Run the app

create-frogbot-app writes a .env file with a generated FROGBOT_SECRET, the connection string for your selected database, and your provider key if you entered one. Otherwise, add the key to .env (for example OPENAI_API_KEY=...). With OpenAI, Anthropic, or Google, add it before starting the app; the app does not start without it. Bedrock and Zen apps start without a key, but chat needs one.
Use the equivalent dev command if you selected pnpm, yarn, or bun. Then open http://localhost:3000. On the first visit, you create the first admin user. The dashboard then shows a card for each collection; open Chats in the sidebar and select Create New to talk to your agent. To change app behavior, edit src/frogbot.config.ts; see Configuration for the full schema. See CLI for every frogbot command.

​
Software requirements

  • Node.js 24.21 or newer
  • TypeScript 6.0.3 or newer
  • npm, pnpm, yarn, or bun
  • SQLite, PostgreSQL, or MongoDB — see Database
  • Optional: an object store (S3, GCS, Azure, R2, or Vercel Blob) — see Configuration

​
Monorepos

Keep one copy of frogbot in a workspace. Each copy has its own types, so a block, field, or collection exported from a shared package fails to typecheck in an app that resolves a different copy. Check which copies are installed:
A shared package lists frogbot under peerDependencies, plus devDependencies so it can build and run its own tests. Never list it under dependencies, which installs a separate copy for that package. Use the same frogbot version in every package. If pnpm why frogbot still shows more than one copy, run pnpm dedupe. An optional peer that only some packages install, such as @ai-sdk/otel, can also split frogbot into two copies; install it in every package that installs frogbot, or in none of them.

​
Next steps

Configuration

Configure app identity, auth, models, pieces, storage, and plugins.

Database

Choose a database adapter and review the core collections FrogBot registers.