Spice AI

Spice AI's page describes Spice as the open data lakehouse for agents. Applications and AI agents use one SQL endpoint to query databases, data lakes, and warehouses, with acceleration into in-memory or on-disk local copies, hybrid search combining keyword, vector, and full-text search in SQL, and LLM inference callable from SQL. A single spicepod.yaml file defines the lakehouse, and the Spice Cloud API can deploy it per customer, team, or agent. Spice runs alongside existing operational databases, which remain the source of truth, and can keep an isolated analytics replica fed by change data capture. Spice is written in Rust and uses Apache DataFusion, Apache Arrow, and Apache Ballista. Agents reach Spice through a SQL API, an OpenAI-compatible chat API, and an MCP server. The page lists over 40 connectors and mentions Twilio, Barracuda Networks, NRC Health, Basis Set, and Summation as production users. Spice Open Source is stated to be free under Apache 2.0; Spice Cloud uses subscription and usage-based plans, and enterprise plans add commercial licensing, 24/7 support, and SLAs.

Key features

• SQL federation and acceleration across operational databases, data lakes, and warehouses
• Analytical query on operational data via isolated analytics replica with CDC
• Hybrid search combining keyword, vector, and full-text search in SQL
• Embedded AI inference: call hosted or local LLMs inline using SQL UDFs or natural language
• Secure AI sandboxing with isolated, least-privilege datasets and row-level filters and column masking per agent identity
• Edge-to-cloud deployment on laptop, on-premises, edge, or cloud, including sidecar, microservice, and cluster topologies
• Tiered storage across memory, local NVMe, and object storage
• Distributed observability with tracing across SQL, embeddings, search, and LLM calls
• Agent access via SQL API, OpenAI-compatible chat API, and MCP server
• 40+ connectors to databases, warehouses, data lakes, and HTTP APIs

Best suited for

• Serving enterprise data to AI agents with isolated, accelerated access
• SaaS companies giving each customer its own lakehouse
• Enterprise data platform teams deploying a lakehouse per internal team
• Running low-latency analytics on operational data without loading production databases
• Grounding AI applications in SQL-queried enterprise data (RAG)

Pricing

Spice.ai Open Source is free under the Apache 2.0 license. Spice Cloud uses subscription and usage-based plans; enterprise plans add commercial licensing, 24/7 support, and SLAs. Specific prices are not stated on the page.