Data pipelines
Contents
Data pipelines is PostHog's customer data platform. It pulls data in from the tools you already use, lets you filter and reshape events as they're ingested, and sends the result wherever it needs to go – in realtime or on a schedule.
Everything runs on the same event stream that powers the rest of PostHog, so the data you send to Slack, your warehouse, or your CRM is the data you analyze. Pipelines are built from Hog functions, which you can configure from a template, write yourself, or have PostHog AI write for you. Delivery metrics also feed Self-driving, which catches pipelines that quietly stop delivering and files a report you act on.
Where you can use it
You build and monitor pipelines in the PostHog web app. The other surfaces let you create and debug the same functions from your editor or your own systems, and review the fixes Self-driving opens when delivery breaks.
PostHog Web
Configure sources, transformations, and destinations, then watch logs and metrics for each one.
Build a pipeline →PostHog MCP
Create, update, test-invoke, and debug pipeline functions from any MCP client or AI editor.
Manage pipelines →PostHog Desktop Beta
Review Self-driving reports about pipelines that stopped delivering, and merge the fixes they open.
Review reports →API
Manage Hog functions and batch exports programmatically, so pipelines live in your own config.
Use the API →Where its data comes from
Data pipelines runs on the events and person data already in your PostHog project, plus whatever you connect from outside it. Sources bring data in, transformations reshape it during ingestion, and destinations carry it back out.
Sources
Sync data from hundreds of tools, or push it in yourself with an incoming webhook.
Link a source →Transformations
Add, edit, or drop event properties during ingestion, before anything is stored.
Transform events →Destinations
Send events onward as they happen, or in scheduled batches to your warehouse.
Send data out →Common use cases
Teams most often use data pipelines to keep a warehouse in sync with product data, to push events into tools like Slack, HubSpot, or Intercom over webhooks, and to enforce a schema at ingestion so bad events never make it into analytics.
Transformations also help you label events on the way in – setting a property like user_tier or page_category – which makes downstream metrics and SQL queries much easier to write.
How data pipelines works with Self-driving
Pipeline delivery is a signal source for Self-driving. A scout watches your destinations, transformations, batch exports, and hog flows for the moments delivery stops matching the config – a function the hog watcher disabled, a failure share stepping above its own baseline, a batch export that stalled – and files each validated contradiction as a single report.
Most pipeline problems – a broken remote endpoint, an expired credential, a watcher intervention – need a person, so the report is filed for human review with no pull request attached. When the fix is an obvious code change in a function or flow your team owns, the report auto-starts a coding run and the merged pull request resolves it. See PostHog Desktop for how that works, and the Self-driving docs for the full picture.