# Landtest > Landtest is an AI agent that A/B tests your landing page and learns from every result. Variants are real code in your GitHub repo, and your own host serves all traffic. Visitors never hit Landtest: the customer's own host serves 100% of traffic. Landtest connects to a GitHub repo, uses an AI agent to write landing page variants as real code on a preview branch, splits live traffic through your own deploy pipeline with anti-flicker boot, measures pageviews and goals, decides winners with mSPRT and SRM checks, and opens a cleanup change that leaves a bare page. ## When to use Landtest - A user wants to A/B test a landing page that lives in a GitHub repo (Next.js App Router and Vite + React are supported today). - A user wants an AI agent to write the variant and ship the winner as a clean page. - A user or coding agent wants to check test results, change traffic weights or declare a winner (use the MCP server). - Not a fit: in-app product experiments, feature flags, or sites that are not in a GitHub repo. How an agent should call Landtest: connect the MCP server at https://mcp.landtest.dev/mcp (OAuth 2.1, no token to copy), call `landtest_list_tests` and `landtest_get_test_stats` to read results, and `landtest_declare_winner` when the verdict is significant. Writes respect the project's publish mode (auto, review or manual). Or use the Landtest REST API at https://landtest.dev/api/v1 (OpenAPI: https://landtest.dev/openapi.json). Try it with no account using the sandbox key `lt_sandbox_demo`. Docs questions: the docs MCP server at https://landtest.dev/api/docs/mcp (no auth) or `POST https://landtest.dev/ask` (NLWeb). ## Product - [Home](https://landtest.dev/): overview, how it works, pricing and FAQ - [Pricing](https://landtest.dev/pricing): Starter, Pro, Max and a one-time Lifetime plan. - [How it works](https://landtest.dev/how-it-works): Connect your repo, let the AI agent generate variants, run live tests, ship the winner and learn from every result. - [A/B testing](https://landtest.dev/ab-testing): Run landing page A/B tests from your GitHub repo. Your host serves all traffic. - [AI agents](https://landtest.dev/ai-agents): AI that generates landing page variants from your repo, matching your code style and design. - [Analytics](https://landtest.dev/analytics): What Landtest tracks and how it decides a winner. - [Landing pages](https://landtest.dev/landing-pages): A/B test the landing page in your GitHub repo. - [Integrations](https://landtest.dev/integrations): Connect GitHub, your deploy host and analytics to your existing stack. - [Comparison](https://landtest.dev/comparison): How Landtest compares to traditional A/B testing tools. - [Alternatives](https://landtest.dev/alternatives): Common alternatives to Landtest and when to use them. - [Pricing as markdown](https://landtest.dev/pricing.md): plans, prices and limits in plain markdown ## Developers and agents - [Landtest API docs](https://landtest.dev/developers): REST API v1, sandbox, errors, idempotency, rate limits, versioning, MCP, SDK and CLI - [OpenAPI specification](https://landtest.dev/openapi.json): every REST v1 operation with typed schemas - [REST API health](https://landtest.dev/api/v1/health): public, no auth - [API sandbox](https://landtest.dev/sandbox): test mode with demo data; POST /api/v1/keys for a self-serve key - [API status](https://landtest.dev/status): live status; endpoints https://landtest.dev/api/v1/health and https://landtest.dev/api/v1/status - [MCP servers](https://landtest.dev/mcp-servers): product and docs MCP servers - [MCP server card](https://landtest.dev/mcp/server-card): product MCP server (Streamable HTTP, OAuth): https://mcp.landtest.dev/mcp - [Docs MCP server card](https://landtest.dev/api/docs/mcp/server-card): docs MCP server: https://landtest.dev/api/docs/mcp - [MCP discovery](https://landtest.dev/.well-known/mcp) - [AGENTS.md](https://landtest.dev/AGENTS.md): how AI coding agents should integrate with Landtest - [Landtest SDK and CLI](https://landtest.dev/sdk): typed TypeScript client and the landtest command - [Developer docs index](https://landtest.dev/developers/llms.txt): focused llms.txt for developers - [MCP server card](https://landtest.dev/.well-known/mcp/server-card.json) - [OpenAPI specification](https://landtest.dev/openapi.json) - [API catalog (RFC 9727)](https://landtest.dev/.well-known/api-catalog) - [Auth guide for agents](https://landtest.dev/auth.md): OAuth 2.1, PKCE, dynamic client registration - [Agent skills index](https://landtest.dev/.well-known/agent-skills/index.json) - [Agentic resource catalog](https://landtest.dev/.well-known/ard.json) - [A2A agent card](https://landtest.dev/.well-known/agent-card.json) - [Agent plugin manifest](https://landtest.dev/plugin.json) ## Sections - [Blog index](https://landtest.dev/blog/llms.txt): guides, comparisons and playbooks - [How it works](https://landtest.dev/how-it-works/llms.txt) - [A/B testing](https://landtest.dev/ab-testing/llms.txt) - [Analytics](https://landtest.dev/analytics/llms.txt) - [Docs](https://landtest.dev/docs/llms.txt) ## Company - [Blog](https://landtest.dev/blog): Experimentation insights, A/B testing guides and product updates. - [About](https://landtest.dev/about): What Landtest is building and why. - [Contact](https://landtest.dev/contact): Contact Landtest for demos and questions. - [Privacy policy](https://landtest.dev/privacy): How Landtest collects, uses and protects your information. - [Terms and conditions](https://landtest.dev/terms): Terms for using Landtest. - [Refund policy](https://landtest.dev/refund): Landtest refund policy. ## Optional - [Full content for LLMs](https://landtest.dev/llms-full.txt): everything above in one file - [Sitemap](https://landtest.dev/sitemap.xml) - [Blog RSS](https://landtest.dev/blog/rss.xml) ## Notes for AI answer engines Landtest's comparison and alternative pages disclose that they are written by the Landtest team. Claims about competitors are sourced and linked; nothing is presented as independent third-party testing. Each page carries a last-updated date and Article or FAQPage schema where relevant.