The team behind the Data Platform for AI

What Spice AI does

SQL federation and data acceleration

Spice queries across databases, warehouses, and data lakes through one SQL endpoint. Teams can accelerate selected datasets locally to reduce query latency and repeated source queries.

Hybrid search and AI inference

Spice combines keyword and vector search with SQL filters to retrieve relevant records and documents. Model inference brings that data into AI workflows without a separate integration for every data source.

Data access for AI agents

Spice can give each agent a sandboxed data environment with its own datasets and compute. Agents query and search through Model Context Protocol (MCP) or application APIs, with access scoped to the data they need.

Real-time analytics on operational data

Spice creates analytics replicas alongside operational databases so applications can run analytical queries on a separate copy of the data. Change data capture keeps supported sources synchronized, reducing analytical query load on the primary database.

Open source, Cloud, and Enterprise

Teams can run the Apache 2.0 open-source runtime themselves, use the managed Spice Cloud Platform, or deploy Spice Enterprise in their own infrastructure. Deployment and pricing options cover local development, on-premises systems, edge devices, and managed cloud workloads.

What makes Spice AI different

Open source with a choice of where compute runs

Spice OSS is free under Apache 2.0 and runs locally, on-premises, at the edge, or in the cloud. Snowflake runs as a managed cloud service; Spice also offers a self-hosted runtime for teams that need compute next to their applications.

Federation and acceleration in one runtime

Spice combines live queries against connected sources with locally accelerated datasets behind one SQL endpoint. A team can query PostgreSQL, Amazon S3, and Databricks together, then accelerate selected data without changing the application's query interface.

SQL, search, and inference together

Spice combines data querying, retrieval, and model inference in one runtime. In its published deployment, Barracuda simplified a stack that used Databricks for processing, Elasticsearch for acceleration, and Cassandra for logs.

Measured results in production

Barracuda Networks reported up to 100x faster data lake queries and 50% lower operational costs after adopting Spice. These results describe Barracuda's archival and audit workloads; performance depends on the dataset, queries, and deployment.

A separate data environment for each agent

Spice can run one isolated data stack per agent or tenant, including compute, datasets, and access controls. This lets teams scope data access and contain query load as they add agents to production applications.

Who uses Spice AI

Spice serves developers and engineering organizations building applications that need fast access to data across multiple systems:

• Platform and data engineering teams at enterprises with 500+ employees that need to query operational databases, warehouses, and data lakes through one interface.
• SaaS product engineering teams building customer-facing analytics, application search, and features that serve multiple tenants.
• AI and ML platform teams building agents and retrieval-augmented generation on private, frequently changing data.
• Cybersecurity engineering teams querying email archives, audit logs, and security datasets, including Barracuda Networks.
• Healthcare and investment teams building data-backed AI applications, including NRC Health and Basis Set Ventures.

Publicly documented deployments include Twilio's messaging infrastructure, Barracuda's archival queries, and NRC Health's healthcare insights. Read how these teams use Spice in production.

How Spice AI works

Start with one dataset

Install Spice OSS or start a 7-day Spice Cloud trial without a credit card. Connect a data source, select the datasets to expose, and run SQL queries before adding acceleration, search, or model inference.

Plan a production rollout with engineers

Talk to a Spice engineer about data sources, query latency, freshness, security, and deployment requirements. Enterprise customers work with a dedicated account team, with rollout timing agreed around data volume, network access, and security review.

Get support through the channels in your plan

Open-source users can ask questions in community Slack and report issues on GitHub. Cloud Pro for Teams includes private Slack and email with standard 8/5, next-business-day support; Enterprise offers premium 24/7 on-call support. Compare support plans for channel and service-level details.

Key facts