Integration · VWO Feature Management & Experimentation

Feature flag CRUD from the IDE.

VWO FME (Feature Management & Experimentation) ships an open-source MCP that runs locally via npx.

The MCP focuses on flag CRUD: rule management, rollout configuration, A/B and multivariate setup, stale flag detection, and SDK code generation. Important scoping note up top: this MCP covers VWO FME only. VWO's visual A/B testing product is a separate surface and is not covered here; for visual testing the team uses the VWO REST API directly.

Official MCPOpen source. Local via npx. VWO_ACCOUNT_ID + VWO_API_KEY env vars.

What you can do via MCP

Example prompts the agent runs.

  • Create a feature flag 'new-onboarding' with three variations.

    Creates the flag in the FME workspace, sets the three variations, returns the SDK snippet for the requested language.

  • Toggle the pricing-page flag on in production.

    Updates the flag's production environment state via the FME MCP and confirms the change took effect.

  • List all flag rules for the homepage A/B test.

    Returns the rule list (audience, rollout, multivariate splits) so the team can review without opening the dashboard.

  • Find unused flags in my codebase.

    Cross-references VWO flag definitions with codebase grep results and produces a stale flag report ordered by age.

  • Generate Node.js SDK integration for the dark-mode flag.

    Returns ready-to-paste Node.js code wired to the flag key, with the right SDK initialization for the team's account.

VWO FME · MCP

Local MCP server invocation. The agent connects with VWO_ACCOUNT_ID and VWO_API_KEY environment variables.

VWO FME MCP
docs
# VWO FME MCP, local server
# Run: VWO_ACCOUNT_ID=12345 VWO_API_KEY=key_abc \
#      npx -y vwo-fme-mcp@latest

mcp.vwo.create_feature_flag({
  name: "new-onboarding",
  key: "new_onboarding",
  variations: [
    { name: "Control", weight: 33 },
    { name: "Variant A", weight: 33 },
    { name: "Variant B", weight: 34 }
  ],
  rollout: { type: "ab", percentage: 100 }
})

# IMPORTANT: This MCP covers FME only.
# Visual A/B testing uses the VWO REST API directly.
One command sample showing how the agent talks to VWO FME. The MCP exposes the platform's primitives as tools; the agent translates the prompt into the right call.

MCP integration

VWO FME MCP server.

Server
npx -y vwo-fme-mcp@latest
Auth
VWO_ACCOUNT_ID + VWO_API_KEY environment variables
Hosting
Local (npx)
Scoping note
FME only. VWO's visual A/B testing product is NOT covered by this MCP. For visual testing, use the VWO REST API directly or the VWO dashboard.
  • Open-source MCP, runs locally via npx
  • Flag rules: rollout, A/B, multivariate, audience targeting
  • Stale flag detection alongside SDK code generation
  • Pairs with the standard VWO REST API for visual testing

VWO FME MCP docs

Visual demonstration

What this looks like in practice.

VWO FME/Feature flag
Experiment

new_onboarding

Three-variant onboarding test, FME-managed

Environment

production

On100% rollout

Targeting rules

  • Type: multivariate (3 variations)
  • Variation weights: 33 / 33 / 34
  • Audience: all visitors
Three-variation FME flag created via the MCP. The agent edits weights and audience without leaving the IDE.
VWO FME· Stale flag audit
2 stale of 3
  • old_search_v1

    Stale

    Last evaluated 78 days ago

    Pre-redesign search; new search is in production

    Remove
  • checkout_button_color

    Stale

    Last evaluated 53 days ago

    Decided experiment, ship variant on; flag still gating code

    Remove
  • new_onboarding

    Active

    Last evaluated 45 minutes ago

    Active experiment

    Keep
Stale flag audit run by the agent. The MCP reports last evaluation timestamps; the agent proposes the cleanup actions.

CLI alternative

REST API for visual testing; MCP for FME.

VWO's visual A/B testing product uses the standard VWO REST API. The FME MCP is the right interface for AI-driven feature flag work; for visual testing the team calls the VWO REST API directly. Knowing which surface owns which workflow is the most common source of integration confusion.

Pairs with these skills

The PM-experimentation skill suite.

This integration pairs with the forthcoming feature-flagging, experiment-design skills. The skill landing pages and SKILL.md sources land in subsequent dispatches; cross-link hyperlinks are added when the skill pages ship.