# Agentwright > Agentwright Starter (version 0.5.0) is a paid Next.js 16 starter kit for developers building SaaS products on Claude agents. One TypeScript codebase with Postgres (Drizzle): a streaming (SSE) agent loop with parallel tool calls, a 12-turn guard and prompt caching; MCP servers (stdio or Streamable HTTP) as tool sources; per-user token and cost metering with a price snapshot on every usage row; monthly token caps and per-plan rate limits; GitHub sign-in with optional Resend magic links; billing through Lemon Squeezy or Polar (checkout and subscription webhooks that set each user's plan); an eval runner with cost per case; license-key delivery with an admin page; tests and CI. Price rows ship for Claude Fable 5.2, Fable 5.1, Opus 5.5, Sonnet 5.5 and Haiku 5.5. Price: $99 at launch for the first 50 licenses (list $149; the code LAUNCH is applied at checkout), paid once, every future version included. 14-day refund, no questions asked. Sold through Polar (merchant of record). Delivery is a zip download and a license key by email; there is no GitHub repository access. Not included: Stripe billing (billing is Lemon Squeezy or Polar), teams, organizations or seats, other model providers, charging for overage, dollar spend caps (the cap is monthly tokens), a job queue, RAG or search, file uploads, an MCP server of its own, Markdown rendering in the chat UI. ## Product - [Agentwright Starter](https://agentwright.app/): what the kit does, pricing, delivery and the questions buyers ask - [What's inside](https://agentwright.app/inside/): every file in the kit with a one-line note, and three source files in full (lib/agent/loop.ts, lib/billing/metering.ts, evals/cases.ts) - [Docs](https://agentwright.app/docs/): the docs that ship with the kit - [Setup](https://agentwright.app/docs/setup/): Postgres, Anthropic key, GitHub sign-in, billing, MCP servers and evals - [Deploy](https://agentwright.app/docs/deploy/): Vercel and Neon step by step, notes for other hosts - [Model prices and price snapshots](https://agentwright.app/docs/price-snapshots/): how each usage row stores its cost and the price version used - [License key delivery](https://agentwright.app/docs/license-delivery/): issuing license keys, signed download links and the admin page - [Billing with Polar](https://agentwright.app/docs/polar/): setup, what each webhook event does, the customer portal, limits and what is tested - [Changelog](https://agentwright.app/changelog/): every release of the kit since 0.1.0 - [Starter kits compared](https://agentwright.app/compare/ai-saas-boilerplates/): Agentwright Starter next to eight other SaaS starter kits on price, refunds, LLM support, usage metering, billing and teams, with a source link for every fact - [Alternatives](https://agentwright.app/alternatives/): one-to-one comparisons with other ways to start an AI product, each with sources, a check date and the cases where the other product is the better choice - [ShipFast alternative for AI agent products](https://agentwright.app/alternatives/shipfast-for-ai-agents/): what a general SaaS boilerplate covers, and the parts an agent product still needs (tool loop, per-user metering, plan caps, evals) - [AnotherWrapper alternative for a Claude-only agent product](https://agentwright.app/alternatives/anotherwrapper-for-claude/): a multi-provider kit with demo apps next to a kit for one Claude agent with per-user cost metering - [Vercel's free Chatbot template or a paid Claude starter kit?](https://agentwright.app/alternatives/vercel-chatbot-template/): what the free template's repository includes, and what you add before you charge for a product built on it - [Claude agent SaaS boilerplate for Next.js: what it has to cover](https://agentwright.app/claude-agent-saas-boilerplate/): the parts beyond login and payments, with the file in the kit and a post with the code for each; the kit uses its own loop on the Anthropic TypeScript SDK, not the Claude Agent SDK - [FAQ](https://agentwright.app/faq/): price, refunds, license, stack, hosting, models, delivery, what is not included - [Refunds and license](https://agentwright.app/legal/): refund policy and license terms ## Free tools - [Claude API cost calculator](https://agentwright.app/tools/claude-api-cost-calculator/): estimate Claude API cost per request, per user and per month for every current model - [Claude API pricing](https://agentwright.app/pricing/claude/): current price per million tokens for every Claude model, with caching, Batch API and worked examples (one page per model) - [Claude model price comparisons](https://agentwright.app/pricing/claude/compare/opus-5-5-vs-sonnet-5-5/): side-by-side prices for every pair of Claude models (Fable, Opus, Sonnet, Haiku) with the cost gap on real workloads - [What a Claude app costs to run](https://agentwright.app/pricing/claude/cost/chatbot/): worked monthly cost of a chatbot, support bot, coding agent, RAG app, document summarization and classification on every Claude model - [Claude Pricing MCP server](https://agentwright.app/tools/claude-pricing-mcp/): free remote MCP server (https://agentwright.app/mcp) with current Claude API prices and cost estimates for Claude, Cursor or any MCP client - [Prompt caching savings calculator](https://agentwright.app/tools/prompt-caching-savings-calculator/): monthly cost with and without Claude prompt caching, savings and break-even calls per cache window - [AI SaaS margin calculator](https://agentwright.app/tools/ai-saas-margin-calculator/): check plan prices against model costs to see gross margin per user ## Blog - [Blog index](https://agentwright.app/blog/): practical guides on building and running Claude agent SaaS (cost metering, tool loops, MCP, evals, rate limits, billing) - [RSS feed](https://agentwright.app/feed.xml) ## Legal - [Terms](https://agentwright.app/terms/) - [Privacy](https://agentwright.app/privacy/)