Statsig MCP: feature flags and experiments in your AI app

Add Statsig to Claude, Cursor or ChatGPT

1. 1Click Connect Statsig free and sign in with your Statsig account.
2. 2Checking gates and configs doesn't log exposures. Logging events or exposures does, permanently, and feeds your metrics, so switch off those logging tools unless you mean your AI app to write analytics data.
3. 3Add ContextGate to your AI app as a connector or MCP server. The steps for each app are on the Works with page.
4. 4Ask your AI app whether a feature gate passes for a test user in Germany, then open Activity to see the check.

Feature flag work from your editor

• Check one feature gate or several for a user context, without logging exposures.
• Get a dynamic config or experiment's value and assignment for a user.
• Read a layer's parameter values, or rank variants for an Autotune config.
• Log custom events or exposures when you do want them recorded.

Keep your experiment data clean

• Switch off event and exposure logging so test calls don't land in your metrics.
• Policies check every call and can hide personal data in user contexts.
• Activity records every gate check and log call your AI apps make.

What you get with Statsig and ContextGate

• Your AI apps can use StatsigClaude, Cursor, ChatGPT and other AI apps reach Statsig through the one ContextGate connection you add to them.
• You choose which tools they can useSwitch off any Statsig tool you don't want used. It's hidden from your AI apps and refused if called.
• Every call is checked and recordedYour policies check each call (for example, personal data is hidden and secrets are blocked), and Activity keeps a record of it.

ContextGate sees only the tool calls that go through it, not your AI app's built-in tools or your chats.

Statsig tools

Examples of what your AI apps can do in Statsig. You'll see the exact list of tools after you connect it.

• Check Feature GateEvaluate one Statsig feature gate with optional user context without logging an exposure. A false result can mean either that the context fails the gate or that the gate name does not exist.
• Check Feature GatesEvaluate multiple Statsig feature gates with optional shared user context without logging exposures. Results are keyed by gate name; a false result does not prove that a gate exists.
• Get Dynamic ConfigEvaluate a Statsig dynamic config or experiment with optional user context and return its value and assignment metadata without logging an exposure. A missing name returns an empty value object.
• Get LayerEvaluate a Statsig layer with optional user context and return its parameter values and assignment metadata without logging an exposure. A missing layer is reported as not found.
• Get Ranked VariantsRank variants for a Statsig contextual-bandit Autotune configuration using optional user context without logging an exposure.
• Log Custom ExposuresPermanently log one or more Statsig gate, experiment, or dynamic-config exposures when automatic exposure logging is not being used. Acknowledged exposures cannot be deleted through the runtime API.
• Log EventsPermanently log one or more custom Statsig events for analytics and metric calculation in a single batch. Acknowledged events cannot be deleted through the runtime API.