Welcome to Spice.ai
Learn what the Spice Cloud Platform does, and find guides for SQL query, AI models, search, data connectors, and SDKs.
The Spice Cloud Platform is a serverless data lakehouse that gives each application and agent fast, sandboxed access to databases, warehouses, data lakes, and APIs. One SQL endpoint retrieves, accelerates, searches, and reasons over operational and analytical data. Applications and agents connect over SQL, an OpenAI-compatible API, or MCP. Text-to-SQL turns questions in natural language into SQL. Spice Cloud is the managed distribution of Spice.ai Open Source, with no infrastructure to manage.
- Get StartedSign up and run your first query in minutes
- SQL Federation and AccelerationQuery databases, warehouses, and lakes with one SQL interface, and accelerate working sets
- Analytics on Operational DataRun analytical queries on a CDC replica of your operational databases
- RAG & Agentic RetrievalCombine SQL filters, vector search, and full-text search in one query, then fuse and rerank the results
- AI InferenceCall models in SQL with
ai(), and get typed answers from decision functions likeai_classify() - Data ConnectorsConnect to 30+ databases, warehouses, and lakes
What You Can Do
Teams use Spice Cloud for:
- Agent data serving: Agents query federated, accelerated datasets with search and tools built in. Each agent sees only the datasets it declares, and Spice holds the source credentials.
- Query across sources without ETL: Run federated SQL across operational databases, warehouses, and APIs. Accelerate the hot path with fewer pipelines to maintain. See Federated SQL Query.
- Real-time analytics on operational data: Run large analytical queries against an analytics replica of your operational database, kept current through high-throughput CDC. See Database Replication and CDC.
- One standard lakehouse for every tenant: Define the lakehouse once in a spicepod. Deploy it as a project for each customer, team, or agent, each with its own sources and settings. See Terraform Provider.
Quick Start
Get up and running in minutes:
Optional, with an OpenAI, Anthropic, or xAI API key: Add an AI model and chat with your data
To work from a coding agent such as Claude Code, Codex, or Cursor, add the Management MCP server at https://api.spice.ai/mcp.
from spicepy import Client
client = Client("YOUR_API_KEY")
reader = client.query(
"SELECT * FROM my_table LIMIT 10"
)
df = reader.read_pandas()
print(df)Community & Support
- Slack: Ask questions and get help from the team at spice.ai/slack.
- GitHub: File issues and contribute at github.com/spiceai/spiceai.
- Enterprise support: Paid plans include priority support with an SLA.
- Help Center: Browse the Help Center for troubleshooting, guides, and FAQs.