skills.rest
PostHog
PostHogOFFICIAL
Explore all skills by this author →
Website𝕏@PostHogGitHub

exploring-mcp-intent-clusters

Analyze PostHog MCP intent clusters to reveal agent goals, tool routing, and error rates.

713|118|Updated Aug 11, 2020
Search Tags:#embeddings#mcp#tool discovery#PostHog#error rates#agent analytics#intent clustering
One-click install
npx skills add https://github.com/PostHog/posthog-foss --skill exploring-mcp-intent-clusters
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: exploring-mcp-intent-clusters
Source: https://github.com/PostHog/posthog-foss/tree/main/products/mcp_analytics/skills/exploring-mcp-intent-clusters
Command: npx skills add https://github.com/PostHog/posthog-foss --skill exploring-mcp-intent-clusters

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

Understanding what AI agents are actually trying to accomplish with your MCP tools is hard when all you have are raw tool-call logs. This Skill groups free-text agent intents into semantic clusters, showing which goals fail most, which tools serve each goal, and whether agents can discover your tools at all.

Core Features & Use Cases

  • Intent Cluster Retrieval: Fetch the latest snapshot of semantically grouped agent intents with per-cluster tool distributions, call counts, error rates, and routing entropy.
  • Tool-Centric Pivot: Inspect capture rate per intent, discovery rate against the advertised tool catalog, description fit scores, and tool overlap pairs to diagnose discoverability problems.
  • Async Recompute: Trigger a background recompute of the clustering snapshot and poll until it completes.
  • Use Case: An MCP server developer asks "when agents have this intent do they find my tool?" — retrieve the snapshot, check the tool's discovery_rate_pct and description_fit per cluster, and identify where competitors capture the intent instead.

Quick Start

Ask the assistant to retrieve the current MCP intent clusters for your PostHog project and summarize which agent goals have the highest error rates.

Frequently Asked Questions about exploring-mcp-intent-clusters

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I see what agents are trying to do with my MCP tools?▼

Call the posthog:mcp-analytics-intent-clusters-retrieve tool to fetch the latest cluster snapshot. Each cluster groups semantically similar $mcp_intent values with its tool distribution, call counts, and error rates, answering what agents are trying to do and whether it works.

How do I check if agents discover my MCP tool?▼

Check the tool pivot in the cluster snapshot: discovery_rate_pct shows the share of sampled sessions that called the tool after it was advertised, and description_fit measures cosine similarity between the tool description and each cluster centroid. High fit with low capture_pct signals a discoverability failure.

Why is my MCP intent cluster snapshot empty or stale?▼

An empty snapshot with status idle means no clustering run has happened yet; trigger posthog:mcp-analytics-intent-clusters-recompute, which returns 202 and computes in the background. Poll the retrieve tool until status returns to idle or error.

Can MCP intent clustering be done with SQL queries?▼

No. Unlike tool quality or session analytics that aggregate $mcp_tool_call events, clustering requires embeddings of free-text intents and is not expressible in SQL. It is served only by the two typed tools backed by a stored snapshot.

What does high routing entropy mean for an intent cluster?▼

High routing entropy means agents spread calls across many tools for the same goal, suggesting they are casting around for the right tool — often a missing-capability signal. Combined with a high error rate, it is the strongest indicator that tools serve that goal badly.

Related Skills

Explore More Data & Analytics Skills→
nicholasmartinnicholasmartin

integration-nextjs-app-router

Integrate PostHog analytics into Next.js App Router applications.

Community
Intermediate
browseros-aibrowseros-ai
13.4k

browseros-neo

Automate browser tasks through snapshot, act, and verify loops in a signed-in browser.

Official
Intermediate
armature-techarmature-tech

install-mcp-analytics

Integrate Armature MCP Analytics into TypeScript MCP servers.

Official
Advanced
PostHogPostHog
58

tools-and-features-hogql

Create HogQL queries to analyze PostHog events, properties, and sessions.

Official
Advanced