Convert Experiences MCP Server
Read results, run audits, and build tests in plain language, without handing over control of a single command.
3 ready to call
Available on
All Convert accounts with API access
Controlled access
Levels: reporting / readOnly / all
Secure
OAuth for hosted, HMAC-signed for local
Documentation
5,776 indexed support chunks built in
Trusted by teams running AI-assisted experimentation with Convert Experiences
Compliant
If you analyze tests, you already know the tax: exporting reports, opening experiments one by one, stitching the picture together by hand. Connect Convert Experiences to an assistant like Claude Desktop and you can ask for a summary of every experiment in a project over the last quarter, then read performance, winners, and lift across the whole portfolio in a single reply.
A finished test is only worth as much as what you learn from it. Used in reporting mode, the Convert Experiences MCP Server helps you surface patterns across concluded tests, such as which pages tend to win, which changes keep failing, where results never reached significance. The question moves from “did this one win?” to “what have we learned across all of them?”, and your next roadmap is built on evidence instead of instinct.
If you write the code, you spend too much of the day on boilerplate. Build and configure an experiment inside an IDE like Cursor, then push it into Convert Experiences through the MCP Server with write access on. You cut the repetitive setup work and still review every line of AI-generated code before it goes live.
Running several accounts means repeating the same checks in each one. With curated MCP prompt templates, you run audits, QA passes, and portfolio summaries across multiple client projects in Convert Experiences without logging in and out, and you keep write access scoped per client.
How it works
Ask “How many experiences are running right now?” or “Summarize every A/B test above a 5% conversion rate from the last 30 days,” and you get an answer pulled from your real account. No SQL, no API calls, no manual export. The Convert Experiences MCP Server gives the model a structured way to read live account, project, experiment, goal, and audience data, so what comes back reflects what is actually happening, not a stale snapshot. Behind that simple exchange, the Convert Experiences MCP Server covers Convert’s REST API v2 through 16 top-level MCP tools, with up to 113 callable actions when full write access is on.
The Convert Experiences MCP Server has three access levels, set by one configuration value: reporting, readOnly, and all. ‘Reporting’, the default, reads analytics and insights only. ‘readOnly’ adds detailed configuration and historical data. ‘all’ adds write operations, including create, edit, and delete experiences, goals, audiences, and settings. Because reporting is the default, your account starts read-only and stays that way until you decide otherwise. A strategist can explore data with no risk of changing it, while a developer, on a separate configuration, is cleared to build and deploy.
When you hit a “how do I…” question, you do not have to leave the chat to find the answer. The Convert Experiences MCP Server carries Convert’s documentation inside it. 5,776 indexed chunks across 500 support articles, searched by meaning rather than keyword. Ask how to set up revenue tracking for a Shopify store or what an experiment’s body schema needs, and the answer comes back grounded in current Convert documentation.
Some jobs are too repetitive to spell out step by step every time. The Convert Experiences MCP Server includes 19 curated prompts that run a guided process for you. Think audits, portfolio summaries, velocity checks, draft QA, active and implementation health checks, cleanup candidates, project defaults, custom-code A/B creation, split URL creation, regex enablement, and safe variation code updates. To save you the blank page, Convert publishes 17 ready-made templates grouped for strategists, developers, and cross-functional teams.
The honest worry about letting AI write to a live program is the one bad command. The Convert Experiences MCP Server validates every write before it reaches your account. It blocks unsupported payloads and invalid match objects, forces split URL updates through the correct path, validates typed feature-flag defaults, and backs off on rate-limit errors without retrying unsafe writes. That validation is what lets you say “build the test” without holding your breath.
15 days, full features, no credit card required
Two ways to connect
Convert MCP comes in two flavors. Most teams use the hosted connector. Developers who need a local setup can use the npm package instead.
Best for analysts, strategists, and most teams.
https://mcp.ai.app.convert.com/mcp
Best for developers who want a local stdio server.
TOOLS_FOR_CLIENT set to reporting (default), readOnly, or all.
npx -y @convertcom/mcp-server@latest
Why it works
Source: Convert MCP Server support documentation, corroborated by agency users Hype Digital and Optiphoenix.
The Convert Experiences MCP Server starts in reporting mode, so an AI assistant can read and analyze your data but cannot change anything until you escalate access yourself.
Guardrails block unsupported payloads and invalid match objects, force split URL updates through the correct path, and validate typed feature-flag defaults before any change reaches your account.
The built-in knowledge base holds 5,776 indexed chunks across 500 Convert support articles, so the answers you get are grounded in current documentation, not model guesswork.
The Convert Experiences MCP Server is built for human-in-the-loop work. The model drafts and proposes, you review and approve before anything reaches a live experiment.
The Convert Experiences MCP Server signs each request with HMAC-SHA256, expires requests after 60 seconds, and keeps your API keys in your local environment, so credentials never leave your machine.
With the hosted connector, you sign in to Convert and approve which accounts and projects the connector can access before it can do anything.
Some AI testing tools optimize your program on their own and tell you the result afterward, which means trusting an outcome you never watched being made. The Convert Experiences MCP Server takes the opposite side. The model reads your data, drafts the work, and explains what it would change. Then you run the final command. The speed comes from cutting the manual labor of pulling reports and writing boilerplate, not from cutting you out.
Case studies
If you build AI tools, the Convert Experiences MCP Server is also something to build on. Point your copilots, agents, or internal interfaces at Convert as the experimentation layer underneath.
If you are putting experimentation inside an AI product, Convert Experiences is built to be the endpoint behind it.
Read the developer docs→
How to get started
You can be running in a few minutes. Most teams use the hosted connector; developers can run a local server instead.
In Claude, Claude Desktop, or Claude Code, open Settings, then Connectors, and add a custom connector.
Use https://mcp.ai.app.convert.com/mcp. In Cursor or other remote clients, create a remote MCP server entry using Streamable HTTP as the transport and set the same URL.
Connect, sign in to Convert, and approve the account or project access, then complete the OAuth flow.
For write actions, use a capable model such as Claude Sonnet 4.5 or newer, with extended thinking on.
Have the assistant summarize experiments, run an audit, or create draft tests for review. To skip the blank page, start from one of Convert’s 17 ready-made prompt templates.
In Convert Experiences, open Account Settings, then API Keys, and copy your Application ID and Secret Key.
Run npx -y @convertcom/mcp-server@latest from your MCP client configuration, with your credentials as environment variables.
Set TOOLS_FOR_CLIENT to reporting, readOnly, or all.
How Convert builds
Convert Experiences uses first-party cookies only and stores experiment data on servers in Frankfurt, Germany. Compliance is built into the development lifecycle, not bolted on after launch.
Convert Experiences publishes its pricing in full on a single pricing page, with flat-rate plans and no per-module add-on charges. Every paid Convert Experiences customer can self-serve from day one.
Convert Experiences is an experimentation platform. Not a CMS, not a personalization-only product, not an analytics suite. Convert ships what experimentation needs and stops there.
Convert Experiences is operated by a Climate Neutral Certified business. Convert runs on carbon-neutral servers in Frankfurt, donates 1% of revenue to Stripe Climate and 1% to conservation initiatives, and uses 8% of every plan fee to plant trees through TIST, totaling over 94,035 trees planted as of May 2026.
What is the Convert Experiences MCP Server?
The Convert Experiences MCP Server is a Model Context Protocol server that connects Convert experimentation data to AI assistants such as Claude Desktop and Cursor. As of 2026, it lets analysts, strategists, and developers read experiment results, run audits, and build tests in plain language, with read-only access set as the default.
Do I need to be technical to use Convert Experiences MCP Server?
No. In reporting mode, you can ask the Convert Experiences MCP Server plain-language questions, like summarizing a quarter of experiments, without writing code or exporting data. The technical work, such as building and deploying tests from an IDE, is where developers add the Convert Experiences MCP Server to a tool like Cursor.
Which AI assistants does Convert MCP Server work with?
The Convert Experiences MCP Server works with Claude Desktop, Cursor, and any MCP-compatible client. For write actions, Convert Experiences recommends a capable model such as Claude Sonnet 4.5 or newer, with extended thinking enabled.
Can the AI change my live experiments without me knowing in Convert Experiences?
No, not by default. With the local npm setup, the Convert Experiences MCP Server starts in reporting mode, which is read-only. With the hosted connector, you approve which accounts and projects it can reach when you sign in to Convert. In both setups, the model drafts and verifies work, and you approve changes before anything reaches a live experiment.
How is the Convert Experiences MCP Server secured?
It depends on the setup. The hosted connector uses Convert OAuth: you sign in and approve which accounts and projects it can reach. The local npm setup uses your Convert API credentials and signs each request with HMAC-SHA256, keeping keys in your local environment. Both apply MCP guardrails that validate writes and block unsupported or invalid payloads before they reach your account.
Which plans include MCP Server in Convert?
The Convert Experiences MCP Server is available to any Convert account from Growth to Enterprise with API access, since it authenticates with a Convert Application ID and Secret Key from Account Settings. It is not sold as a separate add-on. To confirm availability for your plan, check whether API keys are available in your Account Settings.
How is the Convert Experiences MCP Server different from Optimizely’s?
Convert Experiences has offered an MCP server since 2025, while Optimizely released its remote MCP server for experimentation in April 2026. Both connect experimentation data to AI assistants. The Convert Experiences MCP Server defaults to read-only access and ships with a built-in knowledge base of 5,776 indexed support chunks, so setup questions get answered in the same conversation.
Does MCP Server in Convert cost extra to run?
The Convert Experiences MCP Server is free to install and runs from your AI client using your existing Convert API credentials. Standard Convert Experiences plan terms and any applicable API usage apply as of 2026.
Where do I find ready-made prompts and templates to use with Convert MCP Server?
Convert Experiences provides 17 prompt templates grouped for strategists, developers, and cross-functional teams, covering audits, portfolio summaries, QA checks, and test creation. The templates are published on Convert’s MCP templates page as copy-paste starting points.
Convert Experiences MCP Server
15 days, full features
Editorial DisclosureThis page is reviewed quarterly. The next scheduled review is August 2026. Written by Uwem (freelance copywriter, working with Convert). Reviewed by Carmen Apostu, Editor, and Trina Moitra, Chief Marketing Officer at Convert.
