Skip to main content
Welcome to Croissant, the AI-powered product analytics platform that turns raw events into actionable signals.
This guide walks you through getting up and running. If you already track events with a PostHog-compatible SDK, you can plug Croissant in without changing a single line of instrumentation code. Here’s what we’ll cover:
  1. What is Croissant?
  2. Create your account
  3. Generate an API key
  4. Send your first events
  5. See signals appear

​
What is Croissant?

Most analytics tools ask you to define dashboards, write queries, and decide what to look at before anything interesting happens. Croissant works the other way around. You send us raw events — clicks, page views, purchases, signups, API calls, anything — and our backend does the heavy lifting:
  1. Events are ingested and stored.
  2. Insights are automatically generated by analyzing patterns across your event data.
  3. Signals surface the things that actually matter. A signal is a self-explaining, plain-language notification like “Checkout conversion dropped 18% in the last 24 hours” or “Power users in the Enterprise plan are churning faster than usual”.
You don’t need to configure funnels, set thresholds, or build dashboards upfront. Just send events, and Croissant tells you what you should pay attention to.
Croissant is post-hoc compatible — it works with events you already track. If you use a PostHog-compatible SDK, just swap the API host and key. No re-instrumentation required.

​
1. Create your account

Head over to app.trycroissant.ai and sign up. You’ll get an organization created automatically — this is the workspace where your events, insights, and signals live.

​
2. Generate an API key

Once you’re logged in, navigate to Settings and create a new API key. This key is what authenticates your events against your organization.
Your API key is only shown once. Copy it and store it somewhere safe — you won’t be able to retrieve it later.

​
3. Send your first events

Croissant exposes a PostHog-compatible ingestion endpoint. That means if you already use posthog-js, posthog-python, or any PostHog SDK, all you need to do is point it at Croissant.

​
Option A — Using the PostHog JavaScript SDK

If you already have posthog-js installed, just update the host and key: 
import posthog from "posthog-js";

posthog.init("<YOUR_CROISSANT_API_KEY>", {
  api_host: "https://api.trycroissant.ai",
});
That’s it. Every posthog.capture() call now flows into Croissant.

​
Option B — Using a plain HTTP request

You can also send events directly with any HTTP client: 
curl -X POST https://api.trycroissant.ai/i/v0/e/ \
  -H "Content-Type: application/json" \
  -d '{
    "api_key": "<YOUR_CROISSANT_API_KEY>",
    "event": "user_signed_up",
    "distinct_id": "user-42",
    "properties": {
      "plan": "pro",
      "source": "landing_page"
    }
  }'

​
Option C — Batch ingestion

For higher throughput, send events in batches: 
curl -X POST https://api.trycroissant.ai/batch/ \
  -H "Content-Type: application/json" \
  -d '{
    "api_key": "<YOUR_CROISSANT_API_KEY>",
    "batch": [
      {
        "event": "page_viewed",
        "distinct_id": "user-42",
        "properties": { "page": "/pricing" }
      },
      {
        "event": "button_clicked",
        "distinct_id": "user-42",
        "properties": { "button": "upgrade_now" }
      }
    ]
  }'
You don’t need to define event schemas ahead of time. Croissant automatically discovers the structure of your events as they arrive.

​
4. See signals appear

Once events start flowing in, Croissant gets to work behind the scenes:
  1. Event definitions are created automatically — Croissant samples your event payloads, infers their schemas, and catalogs every event type.
  2. Insights are generated by AI agents that analyze your data and build meaningful queries.
  3. Signals are produced when those insights reveal something worth your attention — a trend change, an anomaly, an opportunity.
Open your Croissant dashboard and head to the Signals page. Each signal includes:
  • A title that summarizes what happened in plain language
  • A severity so you know how urgent it is
  • A recommendation on what to do next
  • The evidence backing it up
You can also use the built-in Chat to ask questions about your data in natural language.

​
Key concepts

Events

Raw data points you send to Croissant — user actions, system events, transactions, anything with a name and properties.

Insights

Automatically generated analytical queries that Croissant builds from your event data to track key metrics over time.

Signals

Plain-language, actionable notifications that tell you what changed and why it matters. Think of them as an always-on analyst watching your data.

Chat

Ask questions about your data in natural language. Croissant translates your question into a query and returns an answer.

​
Next steps

Explore the dashboard

See your events, insights, and signals in real time.

Read the full docs

Dive deeper into configuration, customization, and advanced features.
Need help? Reach out to us at hi@trycroissant.ai.