18.6M records/sec ingestion | 6M+ rows/sec queries | Portable Parquet storage

High-Performance Columnar
Analytical Database

Fast ingestion. Fast queries. Built on DuckDB and Parquet.

Use for product analytics, observability, AI agents, IoT, or data warehousing. Single binary. No vendor lock-in.

Get started in seconds

docker run -d -p 8000:8000 \
  -e STORAGE_BACKEND=local \
  -v arc-data:/app/data \
  ghcr.io/basekick-labs/arc:latest

Single binary. No dependencies. Production-ready in minutes.

Get Started ->

Discover Arc Enterprise

240+ deployments

Open source (AGPL-3.0)

Your Data Deserves Better Than Legacy Databases

Billions of events per day. Millions of records per second. Sub-second queries across all of it.

Traditional databases weren't built for this. Arc was.

Real Production Scale:

Product analytics: Billions of user events per day

Observability: Millions of logs, metrics, and traces per second

AI agents: Millions of conversation and context records

IoT & telemetry: Billions of sensor readings across thousands of devices

Built for billion-record workloads.

Why Arc

DuckDB SQL. No Proprietary Query Language.

Not a custom DSL. Not a query language that changes every major version.

Not vendor lock-in through proprietary syntax.

Just DuckDB-powered SQL with window functions, CTEs, and joins.

Track user behavior and conversion funnels

SELECT
  user_id,
  event_type,
  COUNT(*) OVER (
    PARTITION BY user_id
    ORDER BY timestamp
    ROWS BETWEEN UNBOUNDED PRECEDING AND CURRENT ROW
  ) as event_sequence,
  LEAD(event_type) OVER (
    PARTITION BY user_id
    ORDER BY timestamp
  ) as next_event,
  DATEDIFF('second', timestamp, LEAD(timestamp) OVER (
    PARTITION BY user_id ORDER BY timestamp
  )) as time_to_next_event
FROM analytics.events
WHERE timestamp > NOW() - INTERVAL '7 days'
  AND event_type IN ('page_view', 'add_to_cart', 'checkout', 'purchase')
ORDER BY user_id, timestamp;

If you know SQL, you know Arc. Powered by DuckDB.

Window Functions

Moving averages, ranking, and complex aggregations built-in

CTEs & Subqueries

Break down complex analysis into readable, composable parts

JOINs Across Tables

Correlate events, logs, metrics, and any data across tables

Parquet files you actually own

Your data lives in standard Parquet files on S3, MinIO, or local disk.

Arc disappears tomorrow? You still own your data.

Query it with DuckDB, ClickHouse, Snowflake, or any tool that reads Parquet.

Zero proprietary formats. Zero lock-in.

This is what "portable data" actually means.

S3 / MinIO / Local

Store anywhere you want

Standard Parquet

Industry standard format

You Own It

No vendor lock-in, ever

Performance

18.6M records/sec sustained

High-throughput ingestion and analytical queries via MessagePack columnar format.

No degradation. No memory leaks. Just stable performance.

18.6M

records/sec

sustained ingestion

6.29M

rows/sec

query throughput

0.47ms

p50 latency

p99: 3.51ms

Hardware: M3 Max 14-core, 36GB RAM

See Arc in Action

Live demos running on Arc. Real data, real-time ingestion, sub-second queries.

Satellite Tracking

Live

From product analytics to IoT telemetry—Arc handles the scale and performance requirements that matter.

Product Analytics

Track user events, clickstreams, and A/B tests. Query billions of events in seconds with DuckDB SQL. Window functions, CTEs, and complex aggregations out of the box.

User EventsClickstreamsA/B Testing

Observability

Unified storage for logs, metrics, and traces. Fast queries across all telemetry types. 18.6M writes/sec, 6M+ reads/sec. One database instead of three.

LogsMetricsTraces

AI Agent Memory

Store conversation history, context, and embeddings. Columnar architecture for agent memory with fast temporal queries. RAG-ready.

ConversationsContextEmbeddings

IoT & Telemetry

Sensor telemetry, manufacturing data, equipment monitoring. Native MQTT support. Multi-year retention with automatic compaction and 90% storage reduction.

SensorsMQTTTelemetry

Migration