Skip to main content
The fastest way to start is with your coding agent. Firebolt speaks PostgreSQL and these docs are written for agents as well as people, so a tool like Claude Code can scaffold a connection, write the SQL, and load your data from a single prompt. Point it at this page and describe what you want to query. To do it by hand, the path is the same: run an engine, connect with an SDK, and load data.

​
1. Run an engine

  • Local. Install the single binary for a working engine with zero dependencies: 
    bash <(curl -s https://get-core.firebolt.io/)
  • Cluster. Deploy with the Helm chart or the Firebolt Operator.
  • Managed. Start on the managed service with a $200 credit.
Every engine accepts SQL over the PostgreSQL wire protocol and an HTTP endpoint.

​
2. Connect with an SDK

Firebolt ships clients for the languages engineers already use: For raw HTTP and the REST API, see Use the API.

​
3. Load external data

Query files in object storage directly with table-valued functions, with no ingestion step: 
SELECT *
FROM READ_PARQUET(
  URL => 's3://my-bucket/events/*.parquet'
);
READ_CSV, READ_JSON, and READ_AVRO work the same way. See the table-valued functions reference. To make the data resident in Firebolt for the fastest queries, load it with COPY FROM: 
COPY INTO events
FROM 's3://my-bucket/events/'
WITH (TYPE = PARQUET);
See COPY FROM and the Load data guide for credentials, location objects, and incremental loads.

​
Next steps