Firebolt

Firebolt is an analytical database from Firebolt Analytics Inc, aimed at engineers building real-time analytics, ELT, and AI-agent workloads. It uses a Postgres-compatible SQL dialect and can run as a single binary on a laptop or as a cluster of up to hundreds of nodes. The open source edition, Firebolt Core, is described as being in preview, and a managed service is also offered.

What It Is

Firebolt is a distributed analytical query engine built around object storage. Queries are written in a PostgreSQL-compatible SQL dialect with analytical extensions such as array lambda functions, native JSON types, cross-database queries, and geospatial functions. Data can be stored in Firebolt's own format for real-time workloads, or read in place from Parquet and Apache Iceberg for batch use, with DuckLake reading marked experimental. It provides ACID transactions with snapshot isolation.

Deployment Modes

The same engine runs in several ways: a single binary with zero dependencies (SQLite for metadata at laptop scale), a Docker image, a Helm chart, a Kubernetes operator for self-managed clusters, and a fully managed service on AWS. Self-managed deployments are documented as GA on AWS and GCP and in preview on Azure. Firebolt Core is the free self-hosted edition, started with a one-line install script or Docker, and queried via HTTP, a CLI, or a web UI container.

Performance and Indexing

The FAQ describes sparse primary indexes, aggregating indexes, data skipping and full-text indexes, vectorized multi-threaded execution, and tiered caching (FireCache) as the means of keeping latency low. Engines can scale up, scale out, and add clusters automatically for concurrency.

AI and Integrations

SQL functions such as AI_QUERY and AI_EMBED_TEXT call LLMs through Amazon Bedrock, and HNSW indexes with VECTOR_SEARCH provide vector similarity search. A Firebolt MCP Server connects clients such as Claude, Cursor, and GitHub Copilot Chat to Firebolt data, and an in-product Firebolt Agent (preview) helps generate and fix SQL. Ingestion supports COPY FROM, table-valued readers, Kafka streams, and Postgres/MongoDB change data capture (private preview). Drivers exist for several languages, plus dbt, Airflow, Tableau, and Postgres wire protocol connectivity (preview).