PostHog Says 18% of Experiments Were Created via MCP.

PostHog says 18% of experiments created last week were set up through its MCP server, pointing to growing use of AI-assisted workflows for product experimentation. In the LinkedIn post, PostHog describes MCP as shifting experiment creation away from manual engineering work and toward plain-language instructions combined with coding agents.

Key details from the announcement:

• Users can describe a test in plain language instead of manually building the full experiment workflow.
• A coding agent plus PostHog MCP can generate the variant code change.
• The workflow can also create the experiment inside PostHog with a defined success metric.
• PostHog says the setup includes the wiring needed to connect the product change and the experiment tracking.
• The company frames this as moving the main challenge from implementation to deciding which experiments to run.

For small business owners and builders, this could reduce the technical overhead of running product tests, especially for teams without dedicated engineering support. It also suggests a broader shift in no-code and AI tooling, where experiment setup becomes faster but good test design remains the critical skill.