# Agentic A/B testing: what it is and why it's happening now

> AI agents can now do most of the work in A/B testing: find what to test, design and check variants, and watch the results. What differs between tools is whose agent does it, who may press launch, and whether the statistics survive an agent that checks every day.

Last reviewed: 2026-10-17 · HTML version: https://duelpilot.com/agentic-ab-testing

> **Definition:** Agentic A/B testing is experimentation where an AI agent does the research, variant design, QA and monitoring, while a statistics engine decides what won and people approve what goes live.

## Why it's happening now

- [MCP became the common plug.](https://www.anthropic.com/news/model-context-protocol) Anthropic published the Model Context Protocol as an open standard on 25 November 2024. An MCP server exposes tools (read results, propose a test, check a page) that any MCP client can call: Claude, ChatGPT, Cursor and many others. A testing tool with an MCP server can be operated by the agent you already use.
- Agents already do real work. Coding and browser agents edit repositories, fill in CMS screens and check pages in real browsers. Installing a script tag, verifying it and drafting a variant are ordinary agent tasks now.
- Testing is mostly labour. The software is not the expensive part of A/B testing; research, design, QA and analysis are. That labour is exactly what agents are good at, so sites that never had a CRO team can now run tests.
- Velocity beats the one big idea. Many tests don't produce a winner. Cheap, frequent, well-checked tests matter more than a perfect hypothesis, and agents make each extra test cheap.

## Your own agent, or the vendor's agent

There are two models. In the first, the tool exposes an MCP server and [your own agent](https://duelpilot.com/mcp-ab-testing) does the work: it brings your context (your analytics, your repository, your brand notes), and the AI runs on your own subscription. In the second, the vendor runs its own AI inside its product: less setup, but the AI only knows what the vendor's product knows, and it is often billed as credits.

Tools that charge for AI usage, according to their own pages: [GrowthBook](https://duelpilot.com/compare/duelpilot-vs-growthbook) (No credits, but AI usage limits differ by tier (limited on Starter, 5x on Pro, custom on Enterprise)), [Intelligems](https://duelpilot.com/compare/duelpilot-vs-intelligems) ("Unlimited AI usage" on the Unlimited plan; limits on smaller plans not published), [Kameleoon](https://duelpilot.com/compare/duelpilot-vs-kameleoon) (PBX plans include 30, 100 or 200 credits per month, one credit per prompt), [LaunchDarkly](https://duelpilot.com/compare/duelpilot-vs-launchdarkly) (AI runs (LLM tracking/evals): 5K/month included, then $5 per additional 1K runs), [Mida](https://duelpilot.com/compare/duelpilot-vs-mida) (30 MidaGX credits per month on the free plan; unlimited on Growth and Enterprise), [Optimizely](https://duelpilot.com/compare/duelpilot-vs-optimizely) (Agent Platform (Opal) usage incurs credit charges; 200 free credits per month per instance until 31 Dec 2026), [pagent](https://duelpilot.com/compare/duelpilot-vs-pagent) (The license includes a monthly allowance of agentic tasks, with optional task packs on top), [PostHog Experiments](https://duelpilot.com/compare/duelpilot-vs-posthog) (PostHog AI is metered in credits (500 free, worth $5)), [Tailor](https://duelpilot.com/compare/duelpilot-vs-tailor) (No AI credits as such, but plans meter identified-visitor (enrichment) credits), [VWO (Wingify)](https://duelpilot.com/compare/duelpilot-vs-vwo) (Wingz Core is included with paid products; Wingz Advanced (AI Workflow, AI Autopilot, MCP with behavior data) is a paid upgrade) and [Webflow Optimize](https://duelpilot.com/compare/duelpilot-vs-webflow-optimize) (Webflow workspaces include AI credits and sell more; whether Optimize consumes them is not stated).

## Who presses launch: the approval models

The design choice that matters most is who may start a test that real visitors see. An agent that launches by itself is faster; a person who approves catches the variant that is on-brand but wrong, the price claim nobody checked, the test on a checkout page during a sale.

- An agent can launch tests: [ABConvert](https://duelpilot.com/compare/duelpilot-vs-abconvert) (Tools cover launch, pause, resume and end; drafts are created after the user confirms in chat), [Convert](https://duelpilot.com/compare/duelpilot-vs-convert) (Status changes are possible with full access, but activating or pausing requires explicit confirmation; the local server starts read-only), [Eppo (Datadog Experiments)](https://duelpilot.com/compare/duelpilot-vs-eppo) (start\_experiment, conclude\_experiment and cancel\_experiment tools; no approval gate is documented), [Flint](https://duelpilot.com/compare/duelpilot-vs-flint) (publish\_site deploys a page to production; there are no experiment tools), [GrowthBook](https://duelpilot.com/compare/duelpilot-vs-growthbook) (Optional: the API write tool can start experiments; confirmation can be required per API key, and the bundled skills describe draft, review, publish), [Intelligems](https://duelpilot.com/compare/duelpilot-vs-intelligems) (experience\_action can start, pause or end a test; new experiences are drafts and the AI client asks for confirmation before running them), [Kameleoon](https://duelpilot.com/compare/duelpilot-vs-kameleoon) (experiment\_lifecycle\_update can start, pause, resume, stop or delete; only stop and delete require confirmation), [LaunchDarkly](https://duelpilot.com/compare/duelpilot-vs-launchdarkly) (Tools such as toggle-flag, update-rollout and start-experiment-iteration exist; where an environment requires approval, the agent can only file an approval request for a reviewer), [Mida](https://duelpilot.com/compare/duelpilot-vs-mida) (Tests are drafts until launched, but launching and serving the winner are MCP actions with no separate dashboard approval), [pagent](https://duelpilot.com/compare/duelpilot-vs-pagent) (With write scope granted at OAuth the agent can start, pause, stop, roll back and mark winners shipped; pagent adds no second confirmation and relies on the client's confirmation setting), [PostHog Experiments](https://duelpilot.com/compare/duelpilot-vs-posthog) (experiment-launch, -pause, -resume, -end and -ship-variant; approval is left to the MCP client, and write tools can be removed with read-only mode), [Tailor](https://duelpilot.com/compare/duelpilot-vs-tailor) (Optional: "Launch can be open to agents or reserved for humans"; agents can start, pause and ramp experiments when allowed), [Unbounce (Smart Traffic)](https://duelpilot.com/compare/duelpilot-vs-unbounce) (The agent can start and stop tests and deploy winners; A/B testing needs the Experiment plan or higher) and [VWO (Wingify)](https://duelpilot.com/compare/duelpilot-vs-vwo) (With Publish/Admin permission an agent can start, pause or stop a campaign; no server-side approval step is documented).
- A person approves launches: [Coframe](https://duelpilot.com/compare/duelpilot-vs-coframe) (Variations need the customer's approval before going live), [Humblytics](https://duelpilot.com/compare/duelpilot-vs-humblytics) (No start tool; the page says nothing goes live until the user says so and the agent must never promote a winner itself), [Optibase](https://duelpilot.com/compare/duelpilot-vs-optibase) (Read-only MCP) and [Webflow Optimize](https://duelpilot.com/compare/duelpilot-vs-webflow-optimize) (People create and launch in the Webflow UI).

Duelpilot takes the second model on purpose: no MCP tool can approve, launch, resume, pick a winner or ship. An owner can switch on an autopilot policy (which pages, which kinds of change, how much traffic), but that is a setting a person makes, never something an agent can do.

## Statistics that survive an impatient agent

An agent checks results every day. With classic fixed-horizon statistics, looking daily and stopping at the first significant result inflates false positives (the peeking problem). Agentic testing needs statistics that are safe to peek at: sequential, always-valid tests, plus a sample-ratio-mismatch check (a broken split invalidates everything) and guardrail metrics that pause a test that does harm.

Tools that document sequential (peek-safe) statistics: [Amplitude Experiment](https://duelpilot.com/compare/duelpilot-vs-amplitude-experiment) (mSPRT sequential testing is the default), [Convert](https://duelpilot.com/compare/duelpilot-vs-convert) (Sequential frequentist (asymptotic confidence sequences) alongside fixed-horizon and Bayesian), [Eppo (Datadog Experiments)](https://duelpilot.com/compare/duelpilot-vs-eppo) (Sequential frequentist, fixed-sample frequentist, sequential hybrid and Bayesian), [GrowthBook](https://duelpilot.com/compare/duelpilot-vs-growthbook) (Sequential testing available with the frequentist engine; Bayesian is the default), [Kameleoon](https://duelpilot.com/compare/duelpilot-vs-kameleoon) (Sequential testing alongside frequentist (default) and Bayesian), [LaunchDarkly](https://duelpilot.com/compare/duelpilot-vs-launchdarkly) (Sequential testing available with the frequentist method), [Mida](https://duelpilot.com/compare/duelpilot-vs-mida) (Sequential analysis available alongside Bayesian and frequentist), [Optimizely](https://duelpilot.com/compare/duelpilot-vs-optimizely) (Stats Engine is sequential with FDR control; fixed-horizon and Bayesian in beta), [pagent](https://duelpilot.com/compare/duelpilot-vs-pagent) (Bayesian results with lift and uncertainty; sequential testing is not mentioned), [PostHog Experiments](https://duelpilot.com/compare/duelpilot-vs-posthog) (Optional sequential testing on the frequentist engine; Bayesian also available; optional CUPED), [Statsig](https://duelpilot.com/compare/duelpilot-vs-statsig) (Frequentist sequential testing is available, plus a Bayesian mode and CUPED) and [VWO (Wingify)](https://duelpilot.com/compare/duelpilot-vs-vwo) (Bayesian-powered reporting; sequential testing can be enabled to adjust for peeking). Duelpilot uses mSPRT with Holm correction across variants and decides only at full-week boundaries.

## What to look for in an agentic testing tool

1. Whose agent: can you use your own (MCP), or only the vendor's?
2. Who launches: can an agent start a test that visitors see, and can you switch that off?
3. Peek-safe statistics: sequential tests, SRM checks, guardrails.
4. Real-browser checks before launch: phone, tablet and desktop, with screenshots you can see.
5. Copy rules: does anything stop an agent from inventing prices, reviews or claims?
6. Visitor data: what is stored about your visitors, and where.
7. Pricing: flat, per visitor, per seat, or AI credits on top?

## Agentic A/B testing tools compared

Tools with an MCP server (as documented by each vendor), plus Duelpilot. Each cell links to its source.

| Tool | Your agent designs variants | Agent can launch | Sequential stats | AI credits | Pricing model |
|---|---|---|---|---|---|
| Duelpilot | Yes — Your agent designs 2–4 variants per idea, each with a preview and a device check. [source](https://duelpilot.com/agents.md) | No — A person approves every launch; an owner can switch on an autopilot policy with guardrails, never an agent. [source](https://duelpilot.com/index.md) | Yes — mSPRT: always-valid p-values, safe to check daily. [source](https://duelpilot.com/index.md) | No — The AI runs in your own agent's subscription. [source](https://duelpilot.com/pricing.md) | flat per site |
| [Amplitude Experiment](https://duelpilot.com/compare/duelpilot-vs-amplitude-experiment) | Yes — create\_experiment with custom variants, create\_flags, update\_experiment. [source](https://amplitude.com/docs/amplitude-ai/amplitude-mcp) | Unknown — No dedicated start tool documented; update\_flag covers deployments. [source](https://amplitude.com/docs/amplitude-ai/amplitude-mcp) | Yes — mSPRT sequential testing is the default. [source](https://amplitude.com/docs/feature-experiment/under-the-hood/experiment-sequential-testing) | No — AI Agents and MCP included on all plans; agent-analytics sessions are capped per plan. [source](https://amplitude.com/pricing) | by usage |
| [Convert](https://duelpilot.com/compare/duelpilot-vs-convert) | Yes — With full access, the agent can create, clone and update experiences and variations; new work is created as a draft. [source](https://www.convert.com/mcp-server/) | Partly — Status changes are possible with full access, but activating or pausing requires explicit confirmation; the local server starts read-only. [source](https://www.convert.com/mcp-server/) | Yes — Sequential frequentist (asymptotic confidence sequences) alongside fixed-horizon and Bayesian. [source](https://www.convert.com/features/) | No — No AI credits on the pricing page; Convert Intelligence (beta) is listed on all plans. [source](https://www.convert.com/pricing/) | by traffic (visitors or users) |
| [Crazy Egg](https://duelpilot.com/compare/duelpilot-vs-crazy-egg) | Unknown — The documented tools are chat and harness tools; A/B-test actions are not described. [source](https://support.crazyegg.com/knowledge-base/crazy-egg-model-context-protocol-mcp/) | Unknown — No test-launch tool documented. [source](https://support.crazyegg.com/knowledge-base/crazy-egg-model-context-protocol-mcp/) | Unknown — Classic split or bandit; no statistical method named. [source](https://support.crazyegg.com/knowledge-base/crazy-egg-a-b-tests-guide-to-multi-armed-bandit-testing/) | No — AI analysis for heatmaps and recordings included; no credits mentioned. [source](https://www.crazyegg.com/pricing) | by traffic (visitors or users) |
| [Eppo (Datadog Experiments)](https://duelpilot.com/compare/duelpilot-vs-eppo) | Yes — create\_experiment, create\_datadog\_feature\_flag, link\_feature\_flag\_to\_experiment. [source](https://docs.datadoghq.com/mcp_server/tools/) | Yes — start\_experiment, conclude\_experiment and cancel\_experiment tools; no approval gate is documented. [source](https://docs.datadoghq.com/mcp_server/tools/) | Yes — Sequential frequentist, fixed-sample frequentist, sequential hybrid and Bayesian. [source](https://docs.datadoghq.com/experiments/) | Unknown — Not stated for Experiments. [source](https://www.datadoghq.com/pricing/?product=experiments) | by usage |
| [Flint](https://duelpilot.com/compare/duelpilot-vs-flint) | Partly — Claude creates and edits pages, not test variants. [source](https://www.flint.com/docs/claude-mcp) | Partly — publish\_site deploys a page to production; there are no experiment tools. [source](https://www.flint.com/docs/claude-mcp) | Unknown — No statistics engine of its own found. [source](https://www.flint.com/product) | Unknown — Pricing is custom. [source](https://www.flint.com/pricing) | quote only |
| [GrowthBook](https://duelpilot.com/compare/duelpilot-vs-growthbook) | Yes — growthbook\_api\_write can make any write call the REST API allows, including creating experiments. [source](https://docs.growthbook.io/integrations/mcp) | Yes — Optional: the API write tool can start experiments; confirmation can be required per API key, and the bundled skills describe draft, review, publish. [source](https://docs.growthbook.io/integrations/mcp) | Yes — Sequential testing available with the frequentist engine; Bayesian is the default. [source](https://docs.growthbook.io/statistics/overview) | Partly — No credits, but AI usage limits differ by tier (limited on Starter, 5x on Pro, custom on Enterprise). [source](https://www.growthbook.io/pricing) | per seat |
| [Humblytics](https://duelpilot.com/compare/duelpilot-vs-humblytics) | Partly — create\_split\_test drafts a test, but every write is a proposal the user must explicitly approve. [source](https://humblytics.com/skills) | No — No start tool; the page says nothing goes live until the user says so and the agent must never promote a winner itself. [source](https://humblytics.com/skills) | Unknown — Mentions 95% confidence and stop rules; the statistical method is not named. [source](https://humblytics.com) | No — No AI credits or AI tier on the pricing page. [source](https://humblytics.com/pricing) | by usage |
| [Intelligems](https://duelpilot.com/compare/duelpilot-vs-intelligems) | Yes — create\_experience and update\_experience cover variations, targeting and traffic. [source](https://docs.intelligems.io/developer-resources/mcp-server/available-tools.md) | Yes — experience\_action can start, pause or end a test; new experiences are drafts and the AI client asks for confirmation before running them. [source](https://docs.intelligems.io/developer-resources/mcp-server/available-tools.md) | No — Bayesian model with Monte Carlo simulations. [source](https://docs.intelligems.io) | Partly — "Unlimited AI usage" on the Unlimited plan; limits on smaller plans not published. [source](https://www.intelligems.io/pricing) | by usage |
| [Kameleoon](https://duelpilot.com/compare/duelpilot-vs-kameleoon) | Partly — experiment\_create makes a draft with common JS/CSS; no documented tool to create Web variation code. [source](https://docs.kameleoon.com/mcp/mcp-server) | Yes — experiment\_lifecycle\_update can start, pause, resume, stop or delete; only stop and delete require confirmation. [source](https://docs.kameleoon.com/mcp/mcp-server) | Yes — Sequential testing alongside frequentist (default) and Bayesian. [source](https://docs.kameleoon.com/user-manual/experiment-analytics/analyze-results/results-page-settings) | Yes — PBX plans include 30, 100 or 200 credits per month, one credit per prompt. [source](https://www.kameleoon.com/plans) | by traffic (visitors or users) |
| [LaunchDarkly](https://duelpilot.com/compare/duelpilot-vs-launchdarkly) | Yes — create-flag, create-experiment, update-experiment. [source](https://launchdarkly.com/docs/home/getting-started/mcp-hosted) | Yes — Tools such as toggle-flag, update-rollout and start-experiment-iteration exist; where an environment requires approval, the agent can only file an approval request for a reviewer. [source](https://launchdarkly.com/docs/home/getting-started/mcp-hosted) | Yes — Sequential testing available with the frequentist method. [source](https://launchdarkly.com/docs/home/experimentation/analyze) | Partly — AI runs (LLM tracking/evals): 5K/month included, then $5 per additional 1K runs. [source](https://launchdarkly.com/pricing/) | by usage |
| [Mida](https://duelpilot.com/compare/duelpilot-vs-mida) | Yes — Describe the test and Mida builds the variant, targeting, goal and preview link. [source](https://www.mida.so/changelog) | Yes — Tests are drafts until launched, but launching and serving the winner are MCP actions with no separate dashboard approval. [source](https://www.mida.so/mcp) | Yes — Sequential analysis available alongside Bayesian and frequentist. [source](https://www.mida.so/changelog) | Yes — 30 MidaGX credits per month on the free plan; unlimited on Growth and Enterprise. [source](https://www.mida.so/pricing) | by traffic (visitors or users) |
| [Optimizely](https://duelpilot.com/compare/duelpilot-vs-optimizely) | Partly — Can create and update flags, experiments and audiences, but Web variation code (custom HTML, CSS, JS) is not supported through MCP. [source](https://docs.optimizely.com/experimentation-strategy/docs/optimizely-experimentation-mcp-server-overview) | Unknown — No documented start or publish action; the agent confirms in the client before committing changes. [source](https://docs.optimizely.com/experimentation-strategy/docs/optimizely-experimentation-mcp-server-example-prompts) | Yes — Stats Engine is sequential with FDR control; fixed-horizon and Bayesian in beta. [source](https://docs.optimizely.com/experimentation-strategy/docs/statistical-analysis-methods-overview) | Yes — Agent Platform (Opal) usage incurs credit charges; 200 free credits per month per instance until 31 Dec 2026. [source 1](https://docs.optimizely.com/agent-platform/docs/agent-platform-and-ai-features), [source 2](https://docs.optimizely.com/agent-platform/docs/agent-platform-credits) | quote only |
| [pagent](https://duelpilot.com/compare/duelpilot-vs-pagent) | Yes — The agent can capture a hypothesis, turn it into a test and work with pages, audiences and conversion goals. [source](https://pagent.ai/docs/guides/mcp) | Yes — With write scope granted at OAuth the agent can start, pause, stop, roll back and mark winners shipped; pagent adds no second confirmation and relies on the client's confirmation setting. [source](https://pagent.ai/docs/guides/mcp) | Partly — Bayesian results with lift and uncertainty; sequential testing is not mentioned. [source](https://pagent.ai) | Partly — The license includes a monthly allowance of agentic tasks, with optional task packs on top. [source](https://pagent.ai/pricing) | by traffic (visitors or users) |
| [PostHog Experiments](https://duelpilot.com/compare/duelpilot-vs-posthog) | Yes — experiment-create, experiment-create-from-prompt, experiment-update. [source](https://posthog.com/docs/model-context-protocol/tools) | Yes — experiment-launch, -pause, -resume, -end and -ship-variant; approval is left to the MCP client, and write tools can be removed with read-only mode. [source 1](https://posthog.com/docs/model-context-protocol/tools), [source 2](https://posthog.com/docs/model-context-protocol/faq) | Yes — Optional sequential testing on the frequentist engine; Bayesian also available; optional CUPED. [source](https://posthog.com/docs/experiments/statistics-frequentist) | Yes — PostHog AI is metered in credits (500 free, worth $5). [source](https://posthog.com/pricing) | by usage |
| [Statsig](https://duelpilot.com/compare/duelpilot-vs-statsig) | Yes — experiment\_create and gate\_create. [source](https://docs.statsig.com/integrations/mcp) | Unknown — No dedicated start or ship tool is documented; update and destructive tools need a confirmation parameter and review policies still apply. [source](https://docs.statsig.com/integrations/mcp) | Yes — Frequentist sequential testing is available, plus a Bayesian mode and CUPED. [source](https://docs.statsig.com/experiments/advanced-setup/sequential-testing) | Unknown — No AI items on the pricing page. [source](https://www.statsig.com/pricing) | by usage |
| [Tailor](https://duelpilot.com/compare/duelpilot-vs-tailor) | Yes — Agents build experiments and variants: headlines, CTAs, images, full sections, layout and scripts. [source](https://www.tailorhq.ai/integrations/mcp) | Yes — Optional: "Launch can be open to agents or reserved for humans"; agents can start, pause and ramp experiments when allowed. [source](https://www.tailorhq.ai/integrations/mcp) | Unknown — Describes industry-standard methods with a 90% confidence interval; no sequential method named. [source](https://www.tailorhq.ai/features/ab-testing-analytics) | Partly — No AI credits as such, but plans meter identified-visitor (enrichment) credits. [source](https://www.tailorhq.ai/pricing) | by traffic (visitors or users) |
| [Unbounce (Smart Traffic)](https://duelpilot.com/compare/duelpilot-vs-unbounce) | Yes — "Create variants, allocate traffic, start or stop tests, and deploy winning variants." [source](https://unbounce.com/product/mcp-server/) | Yes — The agent can start and stop tests and deploy winners; A/B testing needs the Experiment plan or higher. [source](https://unbounce.com/product/mcp-server/) | No — Smart Traffic is a contextual multi-armed bandit, not a sequential test. [source](https://documentation.unbounce.com/hc/en-us/articles/360036411591-What-is-Smart-Traffic) | No — No AI credits listed; AI copywriting included from Build. [source](https://unbounce.com/pricing) | by traffic (visitors or users) |
| [VWO (Wingify)](https://duelpilot.com/compare/duelpilot-vs-vwo) | Yes — Create A/B and split-URL campaigns, clone and update configuration (not MVT). [source](https://help.wingify.com/hc/en-us/articles/58792565345305) | Yes — With Publish/Admin permission an agent can start, pause or stop a campaign; no server-side approval step is documented. [source](https://help.wingify.com/hc/en-us/articles/58792565345305) | Yes — Bayesian-powered reporting; sequential testing can be enabled to adjust for peeking. [source](https://wingify.com/pricing/) | Partly — Wingz Core is included with paid products; Wingz Advanced (AI Workflow, AI Autopilot, MCP with behavior data) is a paid upgrade. [source](https://wingify.com/pricing/) | quote only |

## Where Duelpilot stands

Duelpilot is built for the own-agent model: a remote MCP server with 48 tools, your agent researches, designs and checks variants, and a person approves every launch. It is a young product: no heatmaps or session recordings, no visual editor for building variants by hand, and fewer integrations than tools that have been around for a decade. If you want a vendor to run optimisation for you end to end, a managed agentic service is the better fit.

## Try Duelpilot with your own agent

Give your agent one line: [set up Duelpilot for my site using /agents.md](https://duelpilot.com/agents.md). It opens a workspace, installs and verifies the script, and drafts the first test; you approve what goes live. [Try your first test free](https://duelpilot.com/login) or [see pricing](https://duelpilot.com/#pricing).

## FAQ

### What is agentic A/B testing?

Experimentation where an AI agent does the research, variant design, QA and monitoring, while a statistics engine decides what won and people approve what goes live.

### Is it safe to let an AI agent run A/B tests?

It can be, if the agent can't launch on its own (or only within strict guardrails), variants are checked in real browsers before launch, copy is checked for invented claims, and the statistics are safe to look at daily.

### Do I need MCP for agentic A/B testing?

Only if you want to use your own agent. MCP is how an agent like Claude, ChatGPT or Cursor connects to a tool. Vendor-run agents don't need it, but then the AI is the vendor's.

### Does agentic A/B testing replace a CRO team?

It replaces much of the labour (research, design, QA, reporting). It does not replace judgement about strategy and brand; that is why a person should still approve what goes live.

## Sources

- [Anthropic: Introducing the Model Context Protocol (25 Nov 2024)](https://www.anthropic.com/news/model-context-protocol) (checked 2026-10-17)
- [Model Context Protocol: What is MCP?](https://modelcontextprotocol.io/introduction) (checked 2026-10-17)
- [Model Context Protocol specification](https://modelcontextprotocol.io/specification/2025-06-18) (checked 2026-10-17)
- [Pricing | Humblytics](https://humblytics.com/pricing) (checked 2026-10-17)
- [Agent-Native CRO for Claude & Codex | Humblytics](https://humblytics.com/skills) (checked 2026-10-17)
- [Humblytics](https://humblytics.com) (checked 2026-10-17)
- [Humblytics MCP](https://mcp.humblytics.com/) (checked 2026-10-17)
- [Pricing | pagent](https://pagent.ai/pricing) (checked 2026-10-17)
- [Pagent MCP server for Claude, ChatGPT, Codex and Cursor](https://pagent.ai/docs/guides/mcp) (checked 2026-10-17)
- [pagent.ai](https://pagent.ai) (checked 2026-10-17)
- [Pricing Plans | Tailor AI](https://www.tailorhq.ai/pricing) (checked 2026-10-17)
- [Tailor MCP: Connect Your AI Agents to Website Experiments](https://www.tailorhq.ai/integrations/mcp) (checked 2026-10-17)
- [A/B Testing Tool for Growth Teams | Tailor AI](https://www.tailorhq.ai/features/ab-testing-analytics) (checked 2026-10-17)
- [Coframe](https://www.coframe.com) (checked 2026-10-17)
- [Agent Platform and AI features | Optimizely](https://docs.optimizely.com/agent-platform/docs/agent-platform-and-ai-features) (checked 2026-10-17)
- [Optimizely Agent Platform credits](https://docs.optimizely.com/agent-platform/docs/agent-platform-credits) (checked 2026-10-17)
- [Optimizely Experimentation MCP server example prompts](https://docs.optimizely.com/experimentation-strategy/docs/optimizely-experimentation-mcp-server-example-prompts) (checked 2026-10-17)
- [Statistical analysis methods overview | Optimizely](https://docs.optimizely.com/experimentation-strategy/docs/statistical-analysis-methods-overview) (checked 2026-10-17)
- [Optimizely Experimentation MCP server overview](https://docs.optimizely.com/experimentation-strategy/docs/optimizely-experimentation-mcp-server-overview) (checked 2026-10-17)
- [Optimizely Experimentation MCP server quickstart](https://docs.optimizely.com/experimentation-strategy/docs/optimizely-experimentation-mcp-server-quickstart) (checked 2026-10-17)
- [Plan page | Kameleoon](https://www.kameleoon.com/plans) (checked 2026-10-17)
- [Kameleoon MCP server](https://docs.kameleoon.com/mcp/mcp-server) (checked 2026-10-17)
- [Results page settings | Kameleoon](https://docs.kameleoon.com/user-manual/experiment-analytics/analyze-results/results-page-settings) (checked 2026-10-17)
- [Connect AI assistants | Kameleoon](https://docs.kameleoon.com/mcp/connect-ai-assistants) (checked 2026-10-17)
- [Mida pricing](https://www.mida.so/pricing) (checked 2026-10-17)
- [Mida MCP server | Connect Mida to your AI agent](https://www.mida.so/mcp) (checked 2026-10-17)
- [Changelog | Mida](https://www.mida.so/changelog) (checked 2026-10-17)
- [Fibr Ai Pricing](https://fibr.ai/pricing) (checked 2026-10-17)
- [Fibr AI Experimentation](https://fibr.ai/experimentation) (checked 2026-10-17)
- [Fibr AI](https://fibr.ai) (checked 2026-10-17)
- [Pricing | Webflow](https://webflow.com/pricing) (checked 2026-10-17)
- [Webflow updates](https://webflow.com/updates/mcp-optimize-analyze-aeo) (checked 2026-10-17)
- [Webflow Optimize](https://webflow.com/feature/optimize) (checked 2026-10-17)
- [Convert Experiences Pricing](https://www.convert.com/pricing/) (checked 2026-10-17)
- [Convert MCP Server](https://www.convert.com/mcp-server/) (checked 2026-10-17)
- [Convert Experiences Features](https://www.convert.com/features/) (checked 2026-10-17)
- [Convert MCP Server | Convert Support](https://support.convert.com/hc/en-us/articles/convert-mcp-server) (checked 2026-10-17)
- [Pricing | Wingify](https://wingify.com/pricing/) (checked 2026-10-17)
- [Connect Wingify with AI Tools using Wingify's MCP Server](https://help.wingify.com/hc/en-us/articles/58792565345305) (checked 2026-10-17)
- [MCP Server | Wingify FME](https://developers.wingify.com/v2/docs/fme-mcp-server) (checked 2026-10-17)
- [AB Tasty Docs](https://docs.abtasty.com/) (checked 2026-10-17)
- [Frequentist analysis mode | AB Tasty Docs](https://docs.abtasty.com/reporting-and-performances/reporting/frequentist-analysis-mode) (checked 2026-10-17)
- [Pricing | PostHog](https://posthog.com/pricing) (checked 2026-10-17)
- [PostHog MCP tools reference](https://posthog.com/docs/model-context-protocol/tools) (checked 2026-10-17)
- [PostHog MCP FAQ](https://posthog.com/docs/model-context-protocol/faq) (checked 2026-10-17)
- [Frequentist statistics | PostHog](https://posthog.com/docs/experiments/statistics-frequentist) (checked 2026-10-17)
- [Pricing & Plans | Intelligems](https://www.intelligems.io/pricing) (checked 2026-10-17)
- [Available Tools | Intelligems Docs](https://docs.intelligems.io/developer-resources/mcp-server/available-tools.md) (checked 2026-10-17)
- [Statistical Significance | Intelligems Docs](https://docs.intelligems.io) (checked 2026-10-17)
- [ABConvert Pricing](https://www.abconvert.io/pricing) (checked 2026-10-17)
- [What Claude can do for you | ABConvert Docs](https://docs.abconvert.io/mcp/tools) (checked 2026-10-17)
- [Drive ABConvert from Claude with MCP](https://docs.abconvert.io/mcp/overview) (checked 2026-10-17)
- [Optibase pricing](https://www.optibase.io/pricing) (checked 2026-10-17)
- [Optibase MCP](https://www.optibase.io/blog/optibase-mcp-ai-assistant-integration) (checked 2026-10-17)
- [Crazy Egg Pricing](https://www.crazyegg.com/pricing) (checked 2026-10-17)
- [Crazy Egg MCP Connector](https://support.crazyegg.com/knowledge-base/crazy-egg-model-context-protocol-mcp/) (checked 2026-10-17)
- [Crazy Egg A/B tests: multi-armed bandit](https://support.crazyegg.com/knowledge-base/crazy-egg-a-b-tests-guide-to-multi-armed-bandit-testing/) (checked 2026-10-17)
- [Unbounce Pricing](https://unbounce.com/pricing) (checked 2026-10-17)
- [Unbounce MCP](https://unbounce.com/product/mcp-server/) (checked 2026-10-17)
- [What is Smart Traffic?](https://documentation.unbounce.com/hc/en-us/articles/360036411591-What-is-Smart-Traffic) (checked 2026-10-17)
- [Pricing | GrowthBook](https://www.growthbook.io/pricing) (checked 2026-10-17)
- [MCP Server | GrowthBook Docs](https://docs.growthbook.io/integrations/mcp) (checked 2026-10-17)
- [Statistics Overview | GrowthBook Docs](https://docs.growthbook.io/statistics/overview) (checked 2026-10-17)
- [Pricing | Statsig](https://www.statsig.com/pricing) (checked 2026-10-17)
- [MCP | Statsig Docs](https://docs.statsig.com/integrations/mcp) (checked 2026-10-17)
- [Frequentist Sequential Testing | Statsig Docs](https://docs.statsig.com/experiments/advanced-setup/sequential-testing) (checked 2026-10-17)
- [Pricing | Datadog Experiments](https://www.datadoghq.com/pricing/?product=experiments) (checked 2026-10-17)
- [Datadog MCP Server Tools](https://docs.datadoghq.com/mcp_server/tools/) (checked 2026-10-17)
- [Experiments | Datadog Docs](https://docs.datadoghq.com/experiments/) (checked 2026-10-17)
- [Pricing | Amplitude](https://amplitude.com/pricing) (checked 2026-10-17)
- [Amplitude MCP](https://amplitude.com/docs/amplitude-ai/amplitude-mcp) (checked 2026-10-17)
- [Sequential testing | Amplitude](https://amplitude.com/docs/feature-experiment/under-the-hood/experiment-sequential-testing) (checked 2026-10-17)
- [Pricing | LaunchDarkly](https://launchdarkly.com/pricing/) (checked 2026-10-17)
- [LaunchDarkly hosted MCP server](https://launchdarkly.com/docs/home/getting-started/mcp-hosted) (checked 2026-10-17)
- [Analyzing experiments | LaunchDarkly](https://launchdarkly.com/docs/home/experimentation/analyze) (checked 2026-10-17)
- [Managed-first pricing | Flint](https://www.flint.com/pricing) (checked 2026-10-17)
- [Claude MCP | Flint Docs](https://www.flint.com/docs/claude-mcp) (checked 2026-10-17)
- [Flint product](https://www.flint.com/product) (checked 2026-10-17)
- [Flint MCP | Connect Claude to Your Landing Page Workflow](https://www.flint.com/product/flint-mcp) (checked 2026-10-17)
- [Duelpilot instructions for AI agents](https://duelpilot.com/agents.md) (checked 2026-10-17)
- [Duelpilot overview](https://duelpilot.com/index.md) (checked 2026-10-17)
- [Duelpilot pricing](https://duelpilot.com/pricing.md) (checked 2026-10-17)

## Related

- [A/B testing MCP servers](https://duelpilot.com/mcp-ab-testing)
- [Best A/B testing tools in 2026](https://duelpilot.com/best-ab-testing-tools)
- [All comparisons](https://duelpilot.com/compare)