Replication Engine: Online // Sub-second CDC

REAL-TIME DATA UNDER 100MS FOR AI APPS

Give your AI features live data, not 15 mins or 1hr old data. Stream every database change to BigQuery under 100ms — no brokers or complex setup.

Single binary · Free forever tier
Pipeline Monitor
3 Active
orders-to-bigquery
RUNNING
postgres → bigquery lag: 72ms
events-to-clickhouse
RUNNING
mysql → snowflake lag: 48ms
users-to-iceberg
RUNNING
mongodb → iceberg lag: 88ms
Throughput
2.4M
events/s
Avg Lag
72ms
end-to-end
Checkpoint
LSN+3
committed
10:42:01WAL decoder: LSN 0/3A1B2C4D received
10:42:01orders INSERT → ArrowRecord[124 rows]
10:42:01AdaptiveBuffer flush triggered (size gate)
10:42:01BigQuery committed: offset 4,821 → 4,945
10:42:02Schema drift check: orders — no changes
10:42:02checkpoint saved: LSN 0/3A1B2C4D
10:42:02WARN: replication lag spike 42ms → resolved
10:42:03users UPDATE → ArrowRecord[87 rows]
10:42:03Iceberg committed: snapshot 8a4f2c
10:42:01WAL decoder: LSN 0/3A1B2C4D received
10:42:01orders INSERT → ArrowRecord[124 rows]
10:42:01AdaptiveBuffer flush triggered (size gate)
10:42:01BigQuery committed: offset 4,821 → 4,945
10:42:02checkpoint saved: LSN 0/3A1B2C4D
vs. competitors: Fivetran: 15 mins to 1hr+ Airbyte: 15 mins to 1hr+
nanosync: <100MS · single binary
// How it works

One binary, stream everywhere

Stream every database change to BigQuery, Kafka, or files in real time. One binary between your database and your destinations — no brokers, no pipelines to maintain.

SOURCE PostgreSQL pgoutput WAL SOURCE SQL Server CDC / tlog NANOSYNC exactly-once · schema-aware SINK BigQuery Storage Write API SINK Kafka Confluent-compatible SINK Files Parquet / CSV / JSONL
72ms
Avg replication lag
1
Binary. No dependencies.
0
Duplicate rows. Ever.
// Development

Ship in Minutes

A single statically linked binary. Drop it anywhere — local, Docker, or Kubernetes. Zero runtime dependencies.

01 / Install

Single binary, any platform

Homebrew
brew install nanosync/tap/nanosync
curl
curl -fsSL https://nanosync.dev/install.sh | sh
Docker
docker pull ghcr.io/nanosyncorg/nanosync:latest
02 / Configure

One YAML file, that's it

nanosync.yaml
connections:
- name: my-postgres
type: postgres
dsn: "postgres://..."
- name: my-bigquery
type: bigquery
pipelines:
- name: orders-to-bq
source:
connection: my-postgres
tables: [public.orders]
sink:
connection: my-bigquery
Test before you run
$ nanosync test connection my-postgres
✓ connected in 12ms
03 / Run

Start and watch live progress

Start
$ nanosync start dev --config nanosync.yaml
✓ nanosync v0.0.4 listening on :7600
Web UI: http://localhost:7600/app
Monitor
$ nanosync monitor
NAME LAG EV/S
orders-to-bq <100ms 4,231
replication live · snapshot complete · no errors
// Reliability

We've Done the Hard Parts
So You Don't Have To

The edge cases that break other pipelines at 2am. We've already solved them.

Exactly-once delivery

No duplicate rows. No silent data loss. Every change lands in BigQuery exactly once — no deduplication logic needed downstream.

Schema drift, handled

Add a column to Postgres? BigQuery gets it on the next batch. Nanosync detects and applies schema changes without restarts or manual intervention.

Checkpoint and resume

Stop the pipeline. Restart it an hour later. It picks up from the exact WAL offset where it stopped — no re-snapshot, no data loss, no manual recovery.

Zero Kafka overhead

No broker to babysit. No consumer group lag to monitor. No topic retention to tune. One binary does what Kafka + Debezium + a connector would — without the operational surface.

Zero impact on production

Logical replication reads the WAL after commits. No query overhead, no table locks, no polling your primary. Your database doesn't know nanosync exists.

Sub-100ms, not "near real-time"

From WAL commit to BigQuery insert in under 100ms. Most tools say "near real-time" and mean minutes. We mean milliseconds — benchmarked, not estimated.

// Pricing

Start free. No connector fees. Ever

Standard

$0 / forever
10M events / month

Prove the latency. All sources and sinks included. Get started in minutes with no credit card.

  • All sources + all sinks
  • 10M events / month
  • Single-node deployment
  • Community support
Download

Business

Custom
 

Multi-node HA, RBAC + SSO, WASM transforms, schema drift approval, 90-day history, custom alerts. Self-hostable, air-gap ready.

  • Everything in Standard
  • Multi-node HA
  • RBAC · SAML 2.0 · OIDC
  • Support with SLAs
Join Waitlist

Enterprise

Custom
 

The autonomous layer. Anomaly detection, natural language pipelines, self-tuning, PII masking, compliance packs, data catalog integration.

  • Everything in Business
  • AI Agents
  • Anomaly detection
  • Self-tuning engine
  • Customer success manager
Join Waitlist