Integrations
Building a graph shouldn't start with an export pipeline.
Point CognoDB at where your data already lives — object storage, a warehouse, a stream — and load it directly from the console's import flow.
Data sources
Load from eleven places.
Amazon S3
CSV and Parquet straight from buckets
Google Cloud Storage
Same-region loads at storage speed
Azure Blob Storage
Container-level batch imports
BigQuery
Query results into the graph
Snowflake
Warehouse tables as nodes and edges
PostgreSQL
Tables and foreign keys → nodes and relationships
Apache Kafka
Streaming ingest with idempotent MERGE
Apache Spark
Bulk transforms feeding batch loads
Delta Lake
Lakehouse tables, versioned reads
Apache Iceberg
Open table format, direct import
CSV & Parquet upload
Drag-and-drop CSV or Parquet
Drivers
The client library is the official one.
CognoDB speaks Bolt 5.0–5.4, so the official Neo4j drivers — Python, JavaScript, Go, Java, .NET — connect without modification. There is no CognoDB SDK to adopt, and none to be locked into.
AI clients
Agents connect over MCP.
The MCP server exposes schema inspection and Cypher to any compatible client.
Idempotent by habit
Loads you can re-run.
Import flows generate MERGE-based writes, so replaying a batch or a stream can't duplicate the graph.
UNWIND $rows AS row
MERGE (c:Customer {id: row.customer_id})
MERGE (p:Product {sku: row.sku})
MERGE (c)-[o:PLACED {order: row.order_id}]->(p)
ON CREATE SET o.at = row.tsStart now
~98.7%
token efficiency at 2,000 entities — see the footnotes above
Bring a bucket, leave with a graph.
The free tier includes the import flow — try it against a real file.
First-graph path
LiveCreate a free instance
No card. Ready in about a minute.
Connect your driver
bolt+ssc:// URI into the driver you already use.
Write two MERGEs
That's the entire shape of agent memory.
Point an agent at it
One MCP config block. No integration code.