PostHog: The Open-Source Platform That Turns Your Product Into a Self-Driving Machine With AI Agents — 36,000+ GitHub Stars

What Is PostHog and Why Is It Trending?

PostHog isn't just another analytics tool. It's the first platform designed around the concept of self-driving products — software that can diagnose its own problems, uncover growth opportunities, and even ship fixes autonomously.

With 36,060 GitHub stars and 437 new stars today, PostHog has become the go-to platform for developers building modern SaaS products who want full control over their product data without vendor lock-in.

The platform captures every signal your product generates — user behavior, errors, performance metrics, feature flag performance — and feeds that context directly to AI agents that can:

• Automatically detect when users encounter friction (rage clicks, failed searches, abandoned flows)
• Generate detailed diagnostic reports with root cause analysis
• Create pull requests to fix identified issues
• Monitor experiment results and alert you to statistically significant changes
• Track AI/LLM performance with built-in observability

The Self-Driving Mode: How It Works

The "self-driving" concept is PostHog's killer feature. Here's what happens when you enable it:

1. Signal Detection

PostHog continuously monitors your product data for patterns that indicate problems or opportunities:

• Error spikes: Sudden increases in JavaScript errors, API failures, or exception rates
• Rage clicks: Users repeatedly clicking elements that don't respond
• Failed queries: Search terms that return zero results
• Drop-off points: Where users abandon multi-step flows
• Performance degradation: Slow page loads, high latency on critical paths

2. Context Aggregation

When a signal is detected, PostHog pulls together all relevant context:

// Example: PostHog captures full session context
{
  "user_id": "user_12345",
  "session_replay": "https://us.posthog.com/replay/abc123",
  "error_stack": "TypeError: Cannot read property 'map' of undefined",
  "page_url": "/checkout/payment",
  "user_actions": [
    "Added item to cart",
    "Proceeded to checkout",
    "Entered shipping info",
    "Clicked 'Pay Now' 3 times",
    "Rage click detected on payment button"
  ],
  "feature_flags": {
    "new_payment_flow": true,
    "express_checkout": false
  },
  "browser_info": "Chrome 115, Windows 11"
}

3. AI Agent Analysis

PostHog's AI agents (or your own agents via MCP) receive this context and can:

• Correlate the error with recent deployments or feature flag changes
• Identify which user segments are affected
• Estimate business impact (revenue at risk, users affected)
• Suggest specific code fixes based on the error pattern

4. Automated Pull Requests

For common issues, PostHog can generate draft pull requests directly. The agent analyzes the stack trace, locates the relevant file in your codebase (via MCP), and proposes a fix with a null-safety check or missing validation — ready for your review.

Complete Product Intelligence Stack

PostHog replaces multiple tools with one integrated platform:

Product Analytics

Autocapture or manually instrument events to understand user behavior. Build funnels, retention charts, paths, and custom queries with SQL or no-code builders.

// Track custom events
posthog.capture('user_signed_up', {
  plan: 'pro',
  referral_source: 'twitter',
  company_size: '10-50'
});

// Identify users
posthog.identify('user_12345', {
  email: 'user@example.com',
  name: 'Jane Developer',
  plan: 'pro'
});

Session Replays

Watch real user sessions to see exactly where they struggle. Filter by user properties, errors, or custom events to find the most relevant sessions. AI agents can automatically flag sessions with the highest frustration signals.

Feature Flags & Experiments

Safely roll out features to specific user cohorts and run A/B tests with statistical rigor:

// Check feature flag
if (posthog.isFeatureEnabled('new-checkout-flow')) {
  showNewCheckout();
} else {
  showLegacyCheckout();
}

// Get experiment variant
const variant = posthog.getFeatureFlag('experiment-pricing-page');
if (variant === 'test') {
  showNewPricing();
}

Error Tracking

Capture frontend and backend errors with full context, group similar errors, and get alerts when error rates spike. Self-driving mode turns these errors into actionable tickets automatically.

AI Observability

Monitor your LLM-powered features with traces, generation logs, latency metrics, and cost tracking:

from posthog import Posthog

posthog = Posthog(api_key='your_key', host='https://us.i.posthog.com')

# Track LLM call
posthog.capture(
    distinct_id='user_123',
    event='llm_generation',
    properties={
        'model': 'gpt-4',
        'prompt_tokens': 150,
        'completion_tokens': 80,
        'latency_ms': 1200,
        'cost_usd': 0.0045
    }
)

Real-World Example: Fixing a Checkout Bug in 10 Minutes

Here's how a team could use PostHog's self-driving mode to catch and fix a critical bug:

9:00 AM: PostHog detects a 300% spike in rage clicks on the payment button for users in the "new-checkout-flow" experiment group.

9:05 AM: AI agent analyzes the pattern and identifies that order.items is undefined when users arrive from the express checkout path.

9:08 AM: Agent generates a pull request with a defensive null check and adds a feature flag to disable the new flow for affected users.

9:10 AM: Developer reviews the PR, confirms the fix, and merges. The flag is toggled off for the broken path until the root cause is addressed.

Total time from detection to fix: 10 minutes. Without PostHog, this bug might have gone unnoticed for days, costing thousands in lost revenue.

Key Benefits for Developers

• Open source & self-hostable: Full control over your data, no vendor lock-in, deploy with one Docker command
• Generous free tier: 1M events, 5k session recordings, 1M feature flag requests, 100k exceptions free every month
• MCP integration: Connect PostHog to Claude Code, Cursor, or any MCP-compatible agent for AI-powered debugging
• Unified platform: Replace 5-7 separate tools (analytics, replays, flags, experiments, error tracking) with one
• SDKs for everything: JavaScript, React, Next.js, Vue, Python, Node, PHP, Ruby, Go, .NET, mobile (iOS, Android, React Native, Flutter)
• Data warehouse sync: Import data from Stripe, HubSpot, or your warehouse and query it alongside product data
• SQL access: Write custom queries when the UI isn't enough — full access to the underlying ClickHouse database
• Transparent pricing: Pay only for what you use beyond the free tier, with clear per-event pricing

Getting Started with PostHog

You can start using PostHog in under 5 minutes:

Option 1: PostHog Cloud (Recommended)

Sign up at us.posthog.com/signup or eu.posthog.com/signup. Install the JavaScript snippet and start capturing events immediately.

Option 2: Self-Hosted Hobby Deploy

Deploy on your own Linux server (4GB+ RAM recommended):

/bin/bash -c "$(curl -fsSL https://raw.githubusercontent.com/posthog/posthog/HEAD/bin/deploy-hobby)"

Option 3: MCP Integration

Connect PostHog to your AI coding assistant so agents can query your product data, diagnose issues, and propose fixes directly from your editor.

Frequently Asked Questions

Is PostHog really free?

Yes, PostHog offers a generous free tier: 1 million events, 5,000 session recordings, 1 million feature flag requests, 100,000 exceptions, and 1,500 survey responses every month. For most startups and small teams, this is more than enough. You only pay when you exceed these limits.

How does PostHog compare to Google Analytics?

PostHog is far more developer-focused. While GA excels at marketing analytics, PostHog provides product analytics, session replays, feature flags, experiments, and error tracking. It also offers full SQL access to your data and can be self-hosted for complete data ownership.

Can I self-host PostHog?

Yes, PostHog is open source and can be self-hosted with a single Docker command. The hobby deployment scales to about 100k events per month. For higher volumes, PostHog recommends their cloud service, but self-hosting is fully supported for smaller deployments.

What is the "self-driving mode" exactly?

Self-driving mode uses AI agents to monitor your product data for signals like error spikes, rage clicks, and failed queries. When detected, the agents generate diagnostic reports and can even create pull requests to fix issues. You remain in control — you review and approve all changes.

Does PostHog work with AI/LLM features?

Absolutely. PostHog has built-in AI observability that tracks LLM calls, including model used, prompt/completion tokens, latency, and cost. You can correlate AI performance with user behavior to optimize your AI features.

How does PostHog handle data privacy?

PostHog offers EU and US cloud regions, supports data residency requirements, and can be self-hosted for complete data control. The platform includes data anonymization options and GDPR-compliant data processing agreements.

Can I migrate from Mixpanel, Amplitude, or Heap?

Yes, PostHog provides migration tools and guides for moving from Mixpanel, Amplitude, Heap, and other analytics platforms. The event structure is similar, making migration straightforward.

Ready to make your product self-driving? Start with PostHog's free tier today and explore our full-stack development courses to build better products faster.