← Back to all sparks
S

Statsig

INFRA · APIS
Velocity3.8

Feature flags, A/B testing, and product analytics platform for engineering and product teams.

Statsig opens an agent-skills repo and broadens MCP, betting on AI agents as a primary user surface.

experimentationfeature flagsproduct analyticsai agentsmcp
Current state
Statsig is treating AI coding agents (Claude, Cursor, Codex) as first-class consumers of the platform. The recent push includes a public agent-skills repository (reusable Statsig skills agents can install), MCP coverage extended to Segments and Layers so targeting and experiment config are now agent-driveable, and a private-beta Console API for generating dashboards programmatically. Alongside that, the analytics surface keeps gaining quality-of-life primitives: cancellable Metrics Explorer queries, Lifecycle Charts, Dashboard Pages.
Where it's heading
The bet is that experimentation, feature flags, and product analytics work increasingly happens through an agent — engineers asking Claude or Cursor to set up a flag, build a dashboard, or check an experiment, rather than navigating Statsig's UI directly. Each release is either widening MCP coverage or making the underlying primitives agent-shaped (Console API for dashboards, shareable metric outputs, query cancellation to keep agent workloads from blowing up warehouse spend).
Prediction
Expect MCP coverage to keep filling out — experiments, gates, holdouts — and the agent-skills repo to become a community surface with sample skills from common LLM agents. Pricing or guardrails around agent-driven warehouse query volume are an increasingly likely follow-on.

Recent moves

  1. 2mo ago

    March 2026 monthly summary page

    Monthly aggregate page ("Mar 2026") rather than a discrete release; appears as scrape noise in the changelog.

    View source ↗
  2. 2mo ago

    Launch of agent-skills repository

    ⚡ SPARK

    Public agent-skills repository launches, giving AI coding agents a curated set of reusable Statsig skills to install. Most directional release in the window — Statsig is treating agents as a primary user surface, not a side integration.

    View source ↗
  3. 2mo ago

    Statsig MCP now supports for both Segments and Layers, so you can more seamlessly manage user targeting and experiment configuration usin…

    Statsig MCP now covers Segments and Layers, so user targeting and experiment configuration are addressable from an agent. Concrete extension of the broader agent-first direction.

    View source ↗
  4. 2mo ago

    Marketing landing page

    Marketing landing page ("Loved by customers at every stage of growth"), not a product release. Appears as scrape noise.

    View source ↗
  5. 2mo ago

    Abort long-running queries from Metrics Explorer to reduce warehouse load and avoid unnecessary compute usage.

    Cancel Queries lets users abort long-running Metrics Explorer queries to avoid runaway warehouse compute. Useful in general and especially as agent-driven query volume grows.

    View source ↗
  6. 3mo ago

    Statsig Agent Skills Repository