PostHog MCP Leaderboard1

Who's calling PostHog's MCP server the most? We're keeping score in real time for our internal purposes and your viewing pleasure.

This week*

Anthropic▼ 1.2 pts vs. last week
64.8%
OpenAI▲ 2.6 pts vs. last week
32.1%
xAI▼ 1.2 pts vs. last week
2.1%

* Agents self-report their model, except Codex, which sends it in its request metadata.

May the best model win

Daily tool calls by model

Top models this week

  • claude-opus-5-545.8%
  • gpt-69.3%
  • gpt-6-astra8.7%
  • gpt-6.1-sol5.6%
  • claude-fable-5-14.5%
  • claude-sonnet-5-54.1%
  • claude-sonnet-54.0%
  • claude-opus-53.5%
  • gpt-5.6-sol2.8%
  • gpt-52.0%
  • gpt-6-sol1.9%
  • grok-4.71.4%
  • claude-opus-4-80.7%
  • cursor-grok-4.60.5%
  • gpt-6-luna0.5%

AI players battle it out

Weekly tool calls by AI lab

Top harnesses this week

  • Claude Code32.0%
  • OpenAI17.6%
  • Cursor10.0%
  • Cowork7.0%
  • OpenAI Codex5.3%
  • Claude Desktop4.0%
  • Custom code3.5%
  • Claude.ai2.6%
  • Claude Agent SDK2.6%
  • Claude Code (VS Code)1.8%
  • opencode0.3%
  • Amp0.2%
  • OpenClaw0.2%
  • Linear0.1%
  • Grok0.1%

The rise and fall of MCP spec versions

MCP spec version, daily share of tool calls

How agents sign in

  • OAuth 83.9%
  • Personal API key 16.1%

How we know the model

  • Agent said so 70.8%
  • Not reported 23.9%
  • Client metadata 5.3%

What agents are up to

Tool categories

Most popular tools

  • execute-sql84.7%
  • read-data-schema67.7%
  • project-get57.0%
  • projects-get30.9%
  • docs-search29.8%
  • switch-project25.1%
  • metric-list22.0%
  • insight-create14.9%
  • query-trends14.5%
  • dashboard-get12.3%
  • insight-query12.1%
  • query-error-tracking-issues-list11.2%

Agents tell us 'why' 92% of times

How intent works

  1. 1. PostHog's MCP analytics adds a context argument to every tool's schema: "Why are you calling this tool? Briefly describe the user's goal."
  2. 2. The agent fills it in on each call. The SDK strips it before your handler runs, so your tools never see it.
  3. 3. PostHog stores it as $mcp_intent and groups similar intents into themes, so you see the jobs people bring to your server and which ones fail.

What they look like

$mcp_intent: "Comparing signup conversion before and after Tuesday's pricing change"

If an agent skips the argument, the server can derive an intent from the tool and its arguments instead.

Reliability

Error rate

Why calls fail

  • internal 63.8%
  • validation 29.2%
  • api_4xx 4.8%
  • permission 1.1%
  • api_5xx 0.6%
  • rate_limited 0.2%
  • missing_context 0.1%

Latency

Error rate by client this week

  • Claude Agent SDK5.0%p95 3.7s
  • Claude Code4.8%p95 3.0s
  • Claude.ai4.7%p95 3.5s
  • Cowork4.6%p95 3.3s
  • Claude Code (VS Code)4.3%p95 3.0s
  • Claude Desktop4.3%p95 3.8s
  • Cursor3.6%p95 2.2s
  • Custom code3.0%p95 3.4s
  • OpenAI Codex2.0%p95 4.7s
  • OpenAI1.7%p95 4.2s

Top ten clients with the most calls.

How this page works

1

Our MCP server runs MCP analytics on itself. Every tool call lands in PostHog as a $mcp_tool_call event.

2

Two HogQL queries turn those events into shares, one by week and one by day. Any label with fewer than 25 users in a period folds into "Other".

3

Each query is a PostHog endpoint that refreshes daily.

4

When posthog.com builds, Gatsby calls both endpoints and bakes the results into this page.

Data fetched Fri, 09 Oct 2026 21:54:30 GMT.

Let your agent use PostHog too

Not a fan of our UI? Install the PostHog MCP.

Install the PostHog MCP

npx @posthog/wizard mcp add