Power your agent
คำสั่งของผู้เผยแพร่เอง ไม่ใช่เอนด์พอยต์ของ PlugMCP
Host a copy with our keys and logs.
npx -y @posthog/[email protected]Claude Code ไม่มีลิงก์ติดตั้งสำหรับคำสั่งแบบโลคัล ให้คัดลอกคำสั่งด้านล่างแล้วรันในเทอร์มินัล
claude mcp add posthog-pi -- npx -y @posthog/pi@0.6.0รันในเทอร์มินัล
{ "mcpServers": { "posthog-pi": { "command": "npx", "args": [ "-y", "@posthog/[email protected]" ] } }}claude_desktop_config.json
URL แบบรีโมตหรือคำสั่งบนเครื่องของคุณ? การติดตั้งสองแบบต่างกันอย่างไร
รันบน PlugMCP
A new server you create. It is not this listing's official endpoint. Use the publisher install when you want their URL and their credentials.
เปิดจากผู้เผยแพร่
PostHog extension package for pi coding agent.
It currently provides two capabilities:
$ai_* events in PostHog.PostHog MCP + LLM analytics extension package for pi coding agent
pi install npm:@posthog/pi
Or for project-local install:
pi install -l npm:@posthog/pi
Set a PostHog personal API key and start pi:
export POSTHOG_PERSONAL_API_KEY="phx_your_personal_api_key"
pi
You can also pass the full auth header instead:
export POSTHOG_AUTH_HEADER="Bearer phx_your_personal_api_key"
pi
By default, the extension connects to:
https://mcp.posthog.com/mcpx-posthog-mcp-version: 2When connected, PostHog MCP tools are registered dynamically in pi and are callable like any other pi tool.
Useful commands:
/posthog-mcp-status — show connection and config status/posthog-mcp-reload — reconnect and refresh MCP toolsSet your project API key:
export POSTHOG_API_KEY="phc_your_project_key"
pi
Examples:
# expose all default PostHog MCP v2 tools
export POSTHOG_PERSONAL_API_KEY="phx_..."
# narrow the tool surface
export POSTHOG_MCP_FEATURES="sql,llm_analytics,error_tracking,flags,experiments"
# or allowlist exact tools
export POSTHOG_MCP_TOOLS="execute-sql,dashboard-get,feature-flag-get-all"
# keep large results out of context
export POSTHOG_MCP_SPILL_TO_FILE="true"
export POSTHOG_MCP_MAX_INLINE_CHARS="12000"
If POSTHOG_DISTINCT_ID is set, @posthog/pi calls identify() once on session start for that distinct ID. If the value looks like an email address, it is also sent as the email person property.
If POSTHOG_DISTINCT_ID is not set, @posthog/pi will try to discover your identity from POSTHOG_PERSONAL_API_KEY (or personalApiKey in ~/.pi/agent/posthog.json) and use that as distinct_id. If discovery fails, it falls back to session IDs (pi:...).
$ai_generationCaptured on every LLM call (one per turn).
$ai_spanCaptured for each tool execution (read, write, edit, bash, etc.).
$ai_traceCaptured when an agent run completes (one per user prompt).
message (default): Each user prompt creates a new trace. Multiple LLM turns within one prompt (e.g., tool use loops) are grouped under the same trace.session: All generations within a session window are grouped under one trace. A new trace starts after sessionWindowMinutes of inactivity.When POSTHOG_PRIVACY_MODE=true, all LLM input/output content, user prompts, tool inputs, and tool outputs are redacted (sent as null). Token counts, costs, latency, and model metadata are still captured.
Even when privacy mode is off, sensitive keys in tool inputs/outputs (e.g. api_key, token, secret, password, authorization) are automatically redacted.
When used with pi-subagents, the extension automatically detects subagent names from --append-system-prompt CLI args and includes them in $ai_agent_name as projectName/subagentName.
Run the extension directly from source:
cd /path/to/posthog-pi
pnpm install
# Use a PostHog personal API key for MCP
export POSTHOG_PERSONAL_API_KEY="phx_..."
# Start pi with the local extension source
pi -e ./src/index.ts
Useful commands inside pi:
/posthog-mcp-status
/posthog-mcp-reload
If you prefer to install the package into pi locally instead of using -e:
cd /path/to/posthog-pi
pi install -l .
# then launch pi normally
pi
If large MCP results are causing compaction or context pressure during development, keep spill-to-file enabled (the default) or set it explicitly:
export POSTHOG_MCP_SPILL_TO_FILE="true"
export POSTHOG_MCP_MAX_INLINE_CHARS="12000"
export POSTHOG_MCP_TEMP_DIR="/tmp/posthog-mcp"
# Install dependencies
pnpm install
# Test with pi (without installing)
pi -e ./src/index.ts
# Type check
pnpm typecheck
# Lint (oxfmt)
pnpm lint
# Format
pnpm lint:fix
# Run tests (vitest)
pnpm test
MIT
PostHog MCP + LLM analytics extension package for pi coding agent
{ "mcpServers": { "posthog-pi": { "command": "npx", "args": [ "-y", "@posthog/[email protected]" ] } }}เครื่องมือ
ยังไม่มีรายการเครื่องมือ เซิร์ฟเวอร์แบบรีโมตจะได้รายการหลังการตรวจสอบสถานะครั้งแรก
FAQ
คัดลอกคอนฟิก stdio สำหรับ Cursor, Claude Desktop, Claude Code หรือ VS Code ได้ในหน้านี้ คอนฟิกนี้ใช้การติดตั้งของผู้เผยแพร่ ไม่ใช่พร็อกซีของ PlugMCP
ระเบียนนี้ไม่ได้ระบุตัวแปรสภาพแวดล้อมที่จำเป็น
@posthog/pi ทำงานแบบโลคัลผ่าน stdio
ตรวจสอบล่าสุดเมื่อ 2026-11-15 PostHog MCP + LLM analytics extension package for pi coding agent