Integrations overview

Integrations forward events that another system already produces into a GraphJSON collection. They are the fastest path for provider data; the logging API remains the better choice for a product-specific event contract.

Open Integrations to connect a provider.

Available integrations#

Each integration writes into a collection, so the same Samples, visualizer, SQL, dashboard, embed, and alert workflows apply.

For another provider, place a custom-source webhook adapter you control between the provider and GraphJSON. The adapter verifies authenticity, minimizes the payload, preserves the source event ID and occurrence time, and handles retries.

For continuous database changes, use change data capture, a transactional outbox, or incremental polling. GraphJSON does not provide a managed database connector; the guide defines the source-reader, mapping, checkpoint, replay, and reconciliation responsibilities for an adapter you operate.

For Kafka, Amazon SQS, Kinesis, Google Pub/Sub, or RabbitMQ, use message-broker and stream-processing ingestion. GraphJSON does not provide managed broker consumers; the guide covers partition-safe checkpoints, acknowledgment, batching, replay, and independent lag monitoring.

For selected OpenTelemetry logs or terminal span outcomes, use OpenTelemetry ingestion and correlation. GraphJSON is not an OTLP or tracing backend; the pattern uses an adapter and keeps the full observability pipeline independent.

For governed aggregate exchange with a warehouse, lakehouse, or BI workflow, use Warehouse and BI handoff patterns. The guide separates bounded Data API extracts, customer-controlled raw dual writes, reverse publication, and direct BI requests.

For Zapier, Make, n8n, Pipedream, and similar tools, use No-code and automation-platform adapters. These are customer-operated HTTP workflows, not managed GraphJSON connectors.

Integration or logging API?#

Choose an integration when:

• the provider already emits the facts you need
• keeping the provider payload is useful
• setup speed matters more than a small, curated schema

Choose the logging API when:

• the event represents a product concept the provider cannot know
• you want a stable, minimal contract
• you need control over identity, names, units, and event time
• the provider payload is too large or sensitive

Many teams use both. Stripe can supply raw billing events while application code emits a small subscription_activated event aligned with product logic.

Choose a custom adapter when the provider signs webhooks, retries deliveries, or emits a payload that must be normalized before GraphJSON receives it.

Choose a database-sync pattern when rows or application transactions are the only dependable source of the fact. Prefer business events from a transactional outbox when you control the writer; use CDC or polling when you must observe an existing database.

Choose a broker consumer when a durable stream already contains the business facts. Acknowledge source positions only after GraphJSON accepts the normalized events.

Choose a telemetry adapter when the source already emits OpenTelemetry but GraphJSON should receive only compact, decision-relevant terminal outcomes.

Choose a warehouse handoff when the consumer needs governed aggregates or the raw source must remain in customer-controlled storage.

Choose an automation platform for low- or moderate-volume workflows with protected secrets, visible failures, bounded retries, and an operational owner.

Plan the collection#

Provider payloads can be wide and change over time. Give each integration a dedicated collection unless its events genuinely share the same retention and analysis workflow with existing data.

Recommended examples:

segment
stripe_events
vercel_production_logs

After connecting:

1. trigger a real test event
2. open Samples
3. identify the fields you actually need
4. set a deliberate retention policy
5. build a small reconciliation query

Then follow Operating managed integrations to define ownership, health signals, schema-drift detection, reconnect behavior, incident response, and a safe disconnect plan.

Security and privacy#

Provider payloads may contain customer or operational data you did not intend to copy.

Review what the source sends, limit event selection at the provider when possible, and avoid forwarding secrets or payment card data. Apply a retention period appropriate to the payload.

Disconnect unused integrations promptly. If an integration-generated destination URL is exposed, contact hi@graphjson.com.