Product

A graph database you provision, not one you operate.

Everything below is running today. Where something is coming later, this page says so rather than implying otherwise.

The query language

Write the traversal, not the joins.

Cypher describes the shape you are looking for. The pattern in the query looks like the pattern in the data.

  • Variable-length and shortest-path traversal
  • MERGE with ON CREATE / ON MATCH semantics
  • Property, composite and full-text indexes
  • Unique, existence and node-key constraints
  • ACID transactions with driver-compatible retry
  • EXPLAIN and PROFILE for plan inspection
Two hops, one querycypher
// Everything a customer touched, two hops out
MATCH (c:Customer {email: $email})-[:PLACED]->(o:Order)
MATCH (o)-[:CONTAINS]->(p:Product)<-[:TAGGED_AS]-(t:Topic)
RETURN t.name AS topic, count(DISTINCT p) AS products
ORDER BY products DESC
LIMIT 10

Capabilities

What you get with every instance.

Cypher, the parts you actually use

Pattern matching, variable-length paths, shortest path, MERGE with ON CREATE and ON MATCH, aggregation and grouping, indexes and constraints, CALL procedures and subqueries, EXPLAIN and PROFILE. A compatibility suite of 2,582 tests runs on every build.

Full-text search beside the graph

Full-text indexes with BM25 ranking live in the same database as your nodes, so a keyword search returns entities you can traverse from immediately — one round trip instead of a search service and a join.

Sizes from 256 MB to 512 GB

Fourteen rungs on the ladder, each with real enforced CPU, memory, connection and storage limits. Move up or down without recreating the instance or changing your connection URI.

Backups that restore to a new instance

Daily snapshots with retention, and restore creates a fresh instance rather than overwriting the original. That makes restore double as fork — clone production into a staging graph without risking it.

Access control on two separate layers

Console and organization access is one thing; the credentials your application uses to open a Bolt session are another. Single sign-on never touches the data path, and instance passwords are shown exactly once.

Automation as a first-class path

Everything the console does is available through the API, an organization-scoped API key, or the CognoDB Terraform provider. Provisioning a database can be part of your signup flow rather than an operations ticket.

Getting data in

Load from where it already lives.

Connectors for object storage, warehouses, streams and table formats — plus CSV and Parquet upload for one-off loads.

Amazon S3Google Cloud StorageAzure Blob StorageBigQuerySnowflakePostgreSQLAmazon S3Google Cloud StorageAzure Blob StorageBigQuerySnowflakePostgreSQLAmazon S3Google Cloud StorageAzure Blob StorageBigQuerySnowflakePostgreSQLAmazon S3Google Cloud StorageAzure Blob StorageBigQuerySnowflakePostgreSQL
Apache KafkaApache SparkDelta LakeApache IcebergCSV & Parquet uploadApache KafkaApache SparkDelta LakeApache IcebergCSV & Parquet uploadApache KafkaApache SparkDelta LakeApache IcebergCSV & Parquet uploadApache KafkaApache SparkDelta LakeApache IcebergCSV & Parquet upload

Where it runs

Three regions.

You choose the region when you create an instance. Data stays in the region you picked.

  • N. Virginia · us-east4
  • Iowa · us-central1
  • Belgium · europe-west1

Architecture

What you connect to.

Your application talks to a managed instance over an encrypted Bolt connection. Sizing, patching, monitoring and backups are ours.

Your applicationNeo4j driverbolt+ssc://encrypted · per-instance credsManaged instancesized · monitored · patchedDaily backupsrestore to a new instance

Sizes

The full ladder, with real limits.

These are the resource limits actually applied to each instance, and prices computed from the same model the console and the billing engine use.

SizeTiervCPUMemoryStorageDisk IOPSConnectionsMonthly
c0freeburst to 0.5 vCPU256 MB1 GBup to 500 IOPS200$0
s1sharedburst to 1 vCPU512 MB1 GBup to 1,000 IOPS256$4.66/mo
s2sharedburst to 1 vCPU2 GB2 GBup to 1,500 IOPS512$10.01/mo
s3sharedburst to 2 vCPU4 GB4 GBup to 2,000 IOPS1,000$19.82/mo
g1shared1 vCPU1 GB2 GBup to 2,000 IOPS512$31.26/mo
g2shared1 vCPU2 GB4 GBup to 3,000 IOPS700$34.97/mo
g4shared1 vCPU4 GB8 GBup to 4,000 IOPS1,000$42.40/mo
g8dedicated2 vCPU8 GB16 GBDedicated1,500$83.31/mo
g16dedicated3 vCPU16 GB32 GBDedicated2,000$139/mo
g24dedicated5 vCPU24 GB48 GBDedicated3,000$221/mo
g32dedicated6 vCPU32 GB64 GBDedicated4,000$277/mo
g48dedicated9 vCPU48 GB96 GBDedicated6,000$414/mo
g64dedicated12 vCPU64 GB128 GBDedicated8,000$552/mo
g128dedicated24 vCPU128 GB256 GBDedicated12,000$1,102/mo
g256dedicated48 vCPU256 GB512 GBDedicated16,000$2,203/mo
g384dedicated72 vCPU384 GB768 GBDedicated20,000$3,303/mo
g512dedicated96 vCPU512 GB1024 GBDedicated24,000$4,404/mo

Storage expands independently

Storage defaults to twice the instance memory and can be expanded on its own up to 4,096 GiB, without a resize.

Not yet available

Multi-region instances, in-place restore and customer-managed encryption keys are not available today. If one of them is a requirement, tell us and we will be straight with you about timing.

Start now

~98.7%

token efficiency at 2,000 entities — see the footnotes above

Put your first graph up in a minute.

A free instance takes about a minute and no card. Write two MERGE statements, read them back, and you have a living graph — with provenance on every fact.

Start free View the quickstart

First-graph path

Live

  1. Create a free instance

    No card. Ready in about a minute.

  2. Connect your driver

    bolt+ssc:// URI into the driver you already use.

  3. Write two MERGEs

    That's the entire shape of agent memory.

  4. Point an agent at it

    One MCP config block. No integration code.