Load data into ClickHouse from 700+ sources

Home

/

Destinations

/

ClickHouse

Airbyte writes to ClickHouse over its native binary protocol with batch inserts. The setup needs a hostname, a port, and a user. No object storage, no staging permissions, no intermediate hop.

Try it freeTalk to sales

How it works

How do you load data into ClickHouse with Airbyte?

Create a user, grant it, point Airbyte at your host. There is no bucket to provision and no storage integration to configure.

01

Create a dedicated user.

Create an airbyte_user rather than reusing an existing one. Set async_insert = 0 on it, which the connector requires for connection checks and syncs to work correctly when async inserts are enabled on the server.
02

Grant permissions.

The user needs CREATE, ALTER, TRUNCATE, INSERT, SELECT, CREATE DATABASE, CREATE TABLE, and DROP TABLE on the target database. Repeat the same grants for any custom namespace.
03

Configure the destination.

Add ClickHouse as a destination in Airbyte and supply the connection details. Then select your streams and run the first sync.
Read the ClickHouse connector docs
Sync pipeline

What happens to your grants when Airbyte syncs?

Not a philosophy. A set of specific, verifiable guarantees about the infrastructure your data runs on.

Why Airbyte

Load ClickHouse without a staging bucket

A dedicated user and network access are enough. Columns stay typed. Writes stay batched.

  • No staging bucket.

    The connector’s requirements are a ClickHouse instance, network access, and a user. Object storage is not among them.

  • Typed columns, not JSON blobs.

    Direct Load writes data into columns matching your source schema rather than storing everything as JSON in raw tables.

  • Batched for throughput.

    Batches are capped at 70 MB, and record window size is tunable if you need to.

Airbyte versus ClickHouse’s native ingestion options

ClickHouse has excellent first-party ingestion for the shapes it was designed around, particularly Kafka and object storage. The gap is everything that starts as a SaaS API.

MethodBest forLimitsUse Airbyte instead when
Kafka table engineStreaming events already on a Kafka topicKafka only. You own the producer and the topicYour data is not on a topic and putting it there is the actual work
s3() and url() table functionsFiles already sitting in object storageYou own extraction, file formatting, and schedulingYour data starts in an API or a database
ClickPipesManaged ingestion on ClickHouse CloudClickHouse Cloud only, and a bounded source listYou are self-hosted, or your source is not covered
PeerDBPostgres change data capture into ClickHousePostgres only. No SaaS connectorsYou need Postgres CDC and SaaS sources in the same cluster
clickhouse-client bulk insertOne-off backfills and manual loadsNot a pipeline. No scheduling, no state, no schema handlingYou need this to keep running after you stop watching it

Where Airbyte wins

Breadth into one cluster. Postgres CDC plus Salesforce, HubSpot, and Zendesk landing in the same ClickHouse database on the same schedule, with schema drift handled, is not something the first-party options cover together.

Where it does not

For a pure Kafka-to-ClickHouse pipeline, the Kafka table engine is simpler.

Raman Singh headshot

Raman Singh

Tech Lead at Symend

"With our legacy framework, if one of the pipelines fails for one client, it will stop everything for the rest of our clients. But with Airbyte, things are run in parallel because of the platform’s distributed nature, which means that we can process multiple clients at the same time without impacting performance."

75%
reduction in sync times
$ 900K
in annual savings
Learn more
Sean Carver, smiling man with beard wearing blue jacket in outdoor setting with warm lighting

Sean Carver

Director of Data at PetDesk

"The real ROI is in our ability to iterate quickly, especially at our increasing scale. At the end of the day, you want a tool like that to just work. We can forget about it and know that it's configured and it's connecting and it's working. That hands-free capability is a big appeal for the platform.”

20+
data sources integrated and growing
+1
FTE engineer in productivity efficiency
85%+
reduction in  data source integration time
Learn more
Man with glasses and goatee smiling at camera in professional headshot

Mondor La Grange

Head of BI and Data Engineering

"Unlike Fivetran's credit-based system that created budget uncertainty, Airbyte's pricing model allows Kuda to forecast expenses accurately and avoid surprise bills."

20+
data sources integrated and growing
+1
FTE engineer in productivity efficiency
85%+
reduction in  data source integration time
Learn more
Smiling woman with glasses and dark hair wearing patterned sweater in shopping mall

Amy Zhao

Senior Manager of Data Engineering

"What's different from Stitch Data or Informatica is the way that we can configure Airbyte connections and Airbyte entities through code. That's a huge plus to us as data engineers, because we are used to checking code and being able to manage changes from Github."

3 to 1
reduction in data integrations solutions for reduced TCO
1
week Shopify and Stripe integration with Airbyte
Learn more
Franziska Ibscher, woman with shoulder-length brown hair and white top, smiling at camera

Franziska Ibscher

Product Manager at Drivepoint

"Airbyte allows us to stay flexible while scaling from hundred-million to billion-dollar enterprise clients."

75%
of customers increased profitability
6.7%
EBITDA increase for customers
Learn more

Which sources can you load into ClickHouse?

All 700+ of them. Most ClickHouse users arrive from Postgres.

Databases

PostgresDatabasesMySQLDatabasesSQL ServerDatabasesMongoDBDatabasesOracleDatabasesDynamoDBDatabases

SaaS applications

SalesforceSaaS applicationsHubSpotSaaS applicationsZendeskSaaS applicationsStripeSaaS applicationsGoogle AdsSaaS applicationsAmplitudeSaaS applications

Files and object storage

Amazon S3Files and object storageGoogle Cloud StorageFiles and object storageSFTPFiles and object storage
Browse all 700+ sources

How does your data land in ClickHouse?

Airbyte writes typed columns directly, and delegates deduplication to ClickHouse’s own engine rather than running MERGE queries against your cluster.

Airbyte typeClickHouse type
StringString
IntegerInt64
DecimalDecimal(38, 9)
BooleanBool
TimestampDateTime64(3)
ObjectJSON if you enable JSON in the connector configuration, otherwise String
ArrayString
UnionString

Schema evolution

  • Column added.

    Airbyte adds the column to the destination table automatically.

  • Type changed.

    Airbyte modifies the column type as needed.

  • Column removed.

    ClickHouse is the exception. Airbyte drops the column and its historical data from the destination table, because ReplacingMergeTree deduplicates by deleting and re-inserting records rather than using MERGE, and that table recreation requires dropping obsolete columns.

The numbers

Teams running Airbyte in production

  • 0M

    pipelines synced daily

  • 18K+

    companies

  • 199%

    ROI, certified by Forrester

Deployment

Where Airbyte runs

Your warehouse credentials and your data do not have to leave your environment. Airbyte does not retain customer data on its servers.

  • Airbyte Cloud

    Fully managed hosting on the Standard and Pro plans. Fastest path to a running pipeline.

  • Airbyte Flex

    Hybrid deployment, with Airbyte’s data plane running in your cloud, VPC, or on-prem environment. You get 700+ connectors and a fully managed experience, while your data never touches our servers.

  • Self-Managed

    Open source, your infrastructure, your rules.

See hybrid deployment

compliance

Enterprise ready

Compliant with standards

SOC 2 Type II certified, GDPR and HIPAA support, with tools to help you meet internal and external regulatory requirements.

Uptime & SLA Guarantees

24/7 support and 99.9% availability backed by contractual SLAs and priority response times for mission-critical workloads.

Flexible deployment options

Move data into your warehouse/lakehouse of choice, VPC, on-prem, or into Airbyte's Context Store.

talk to usTrust Center

Frequently asked questions

Didn’t find your answer?
Please don’t hesitate to reach out.

Talk to us

If events are already on a Kafka topic, the Kafka table engine is simpler. If files already sit in object storage, s3() and url() table functions are enough. ClickPipes is managed ingestion on ClickHouse Cloud with a bounded source list. PeerDB is Postgres CDC only. If you need Postgres CDC plus Salesforce, HubSpot, and Zendesk in the same cluster on the same schedule, Airbyte covers that together.

No. Airbyte writes over the native binary protocol. The setup needs a hostname, a port, and a user. There is no bucket to provision and no storage integration to configure.

Airbyte uses ReplacingMergeTree, so deduplication happens during ClickHouse’s background merges rather than at write time. To guarantee deduplicated results at query time, add the FINAL operator: SELECT * FROM your_table FINAL. Without it you may see duplicated or deleted rows.

ClickHouse is the exception. Airbyte drops the column and its historical data from the destination table, because ReplacingMergeTree deduplicates by deleting and re-inserting records rather than using MERGE, and that table recreation requires dropping obsolete columns.

Pick a monotonically increasing column the source exposes, such as updated_at or an incrementing id, so Airbyte can request only new or changed rows.

The dedicated user needs CREATE, ALTER, TRUNCATE, INSERT, SELECT, CREATE DATABASE, CREATE TABLE, and DROP TABLE on the target database. Repeat the same grants for any custom namespace.

Yes. Airbyte works with self-hosted ClickHouse. Cloud, Flex, Self-Managed, and PyAirbyte are all supported. Flex and Self-Managed keep the data plane in your environment.

Other destinations

Snowflake

Airbyte clones tables instead of swapping them, so the grants you configured survive every sync. Removed source columns are kept, and oversized values are logged rather than fatal.

Load data into Snowflake

BigQuery

Airbyte stages files in Google Cloud Storage, then loads them into BigQuery with native load jobs. Warehouse compute stays on Google Cloud, and Airbyte never sits in the query path.

Load data into BigQuery

Databricks

Airbyte stages Avro files inside Unity Catalog Volumes, not an external bucket, so your data stays under your catalog’s governance from the first byte to the final table.

Load data into Databricks