Get started

Make your new Growth PM

Press to show the next assistant.

Install the A/BLint MCP and flag inconsistencies, missing metrics, and anti-patterns across your A/B test backlog.

Add to Claude
Try our live demo

Built by growth engineers from

Try the real-time linter

hypothesis.md2 errors, 4 warnings

Shorten first-run setup to lift activation

Hypothesis

If we , , because less setup time leaves more room for testing.

Research

Success metric

Guardrails

, and make sure mobile setup doesn’t slip.

Rollout

, then half, then everyone. Stop if activation falls by more than three points in any cohort.

How it works

Install the connector once, then keep working in whatever assistant you already use.

  1. Install the MCP in Claude, ChatGPT, or Grok Bot

    Paste the command in Claude Code, or add https://explint.com/mcp as a connector in settings. Cursor and Gemini read the same server URL.

  2. Ask your agent to lint your test backlog

    Point it at a backlog or a folder of specs. Every finding comes back with the rule it broke and the line it broke on.

  3. Iterate on your A/B test specs

    Ask for a rewrite and lint again until the hypothesis names a metric, a direction, and a reason.

Tools a growth team needs to scale

Set up custom lint rules per project, team, or feature. Use A/BLint wherever you’re working: MCP, API, or Analytics.

Custom lint rules

Toggle the built-in rules on or off, and add rules you want your agents to follow.

API

Send a spec over REST and get the findings back as JSON, with the rule and the line that broke it.

  • One request per spec
  • Findings carry rule and line

MCP

Add explint.com/mcp to Claude, ChatGPT, or Cursor, and the assistant calls the linter for you.

  • One server URL, no key
  • MCP tools in every assistant

Analytics

Quality scores accumulate with every lint, so a weakness that keeps appearing across specifications becomes visible as a pattern the team can fix once.

  • Understand recurring quality issues across linted A/B test specifications
  • Catch drift between experiment platform outcomes and what was agreed at launch
  • Platform-agnostic quality assurance for experimentation programmes

Projects

Give each project its own rules, whether it’s a feature, a team, or a client.

Start free, pay for what you lint

Try it with 150 free lint requests and continue at $0.50 per 1,000 requests thereafter.

Free

$0/ month

  • Up to 150 lint requests
  • Up to one project
  • MCP/API support
Add to Claude

Pay As You Go

$0.50/ 1,000 requests

  • Unlimited lint requests
  • Unlimited projects
  • MCP/API support
  • Custom rules
Create account

A lint request is only counted if the lint tool or endpoint is called. All other requests such as to edit rules or create projects are free.

Questions

What counts as a lint request?

One pass of the linter over one spec. Editing a spec and linting it again is a second request.

Which assistants can install it?

Claude, ChatGPT, Grok Bot, Cursor, and Gemini all work. Claude takes a one-line command, and everywhere else you paste the server URL as a connector or MCP server.

Do I need to learn the API?

There is nothing to learn. The MCP exposes the linter as tools your assistant calls in plain language, and the API is there when you want the same rules inside your own app.

Can I add my own rules?

You can toggle the built-in hypothesis rules on or off, and add rules for the mistakes your team keeps catching by hand.

What does the linter check?

The linter checks your spec against the rules in your project, covering hypothesis, research, design, and metric setup.

Can I try it without an account?

Try the live demo above. To use it from Claude or ChatGPT, create an account to install the MCP.

Don’t let a bad test make it to production

Add A/BLint to your assistant and point it at the backlog.

Add to Claude
Create an account