Kumiki A/B
MCP-first A/B testing server that enables agents to manage experiments (create, update traffic splits, read results, apply winning variants) from tools like Claude Code, with self-hosted Cloudflare backend.
README
Kumiki A/B
Agent-native A/B testing. Free, self-hosted, MCP-first.
Run experiments from Claude Code. Apply winning variants to 100% of traffic — no paywall, no SaaS bill.
Quickstart
# 1. Deploy your backend
npm create kumiki@latest ab
# Follow printed steps: wrangler d1 create → paste id → migrations apply → wrangler deploy
# 2. Add the snippet to your site's <head>
# <script src="https://<your-worker>.workers.dev/s.js?site=SITE_ID"></script>
# 3. Mark conversions
# window.KUMIKI.track('purchase', { value: 4980 })
→ Full install guide — CLI, Deploy-to-Cloudflare button, agent-native, GTM (3-tier), and marking conversions (Step 3).
Why
Every A/B tool charges you for the one thing that actually matters: applying the winner.
VWO, Optimizely, and Optimize Next all gate "winning variant → 100% rollout" behind paid plans (¥100k+/mo). Kumiki A/B makes it free, self-hosted, and fully operable by an AI agent via MCP.
Hero features
- Win-pattern application — free. Roll the winning variant to 100% of traffic with a single MCP call.
- MCP-first. Every operation — create test, update traffic split, read results, apply winner — is callable from Claude Code. No dashboard required.
- Self-hosted on Cloudflare. Experiment data is self-collected into your own D1 — never pulled from a third party. You don't even need GA4 configured to run kumiki; if you have it, we emit events out to it. Zero vendor lock-in.
Stack
| Layer | Tech |
|---|---|
| Snippet | Lightweight JS (<head>) — anti-flicker, sticky bucketing, fail-open, self-collected exposure/conversion events (emitted to GA4 too) |
| API | Cloudflare Workers + Hono + D1 |
| MCP server | Full REST coverage for Claude Code |
| Dashboard | Next.js 15 + visual editor (iframe) |
| Stats | Bayesian (beta-binomial) — peeking-resistant |
| Events | Self-collected store in your own D1 — user-based, windowed; emitted out to GA4/webhook, never pulled |
Status
Early development. Dogfooded on real e-commerce traffic.
License
MIT
Recommended Servers
playwright-mcp
A Model Context Protocol server that enables LLMs to interact with web pages through structured accessibility snapshots without requiring vision models or screenshots.
Magic Component Platform (MCP)
An AI-powered tool that generates modern UI components from natural language descriptions, integrating with popular IDEs to streamline UI development workflow.
Audiense Insights MCP Server
Enables interaction with Audiense Insights accounts via the Model Context Protocol, facilitating the extraction and analysis of marketing insights and audience data including demographics, behavior, and influencer engagement.
VeyraX MCP
Single MCP tool to connect all your favorite tools: Gmail, Calendar and 40 more.
graphlit-mcp-server
The Model Context Protocol (MCP) Server enables integration between MCP clients and the Graphlit service. Ingest anything from Slack to Gmail to podcast feeds, in addition to web crawling, into a Graphlit project - and then retrieve relevant contents from the MCP client.
Kagi MCP Server
An MCP server that integrates Kagi search capabilities with Claude AI, enabling Claude to perform real-time web searches when answering questions that require up-to-date information.
E2B
Using MCP to run code via e2b.
Neon Database
MCP server for interacting with Neon Management API and databases
Exa Search
A Model Context Protocol (MCP) server lets AI assistants like Claude use the Exa AI Search API for web searches. This setup allows AI models to get real-time web information in a safe and controlled way.
Qdrant Server
This repository is an example of how to create a MCP server for Qdrant, a vector search engine.