FLUID
One declarative contract for trustworthy, governable, agent-ready data products.
Contract-First
One .fluid.yml is the source of truth — schema, quality, build, lineage, governance. Version-controlled and schema-validated.
Agentic-Native
agentPolicy, sovereignty, and semantics give LLM agents deterministic answers to PII, allowed-use, metric, and residency questions — enforced at the MCP gateway.
Operational Superset
The only open spec covering build + orchestration + source-aligned acquisition alongside contract and governance.
Interoperable
Compiles to Bitol ODPS + ODCS via the forge-cli reference compiler for DataHub / OpenMetadata / Datamesh Manager catalog interop.
Federated by Design
Built for Data Mesh — decentralized ownership, globally unique product ids, one unified fabric.
Open & MIT-Licensed
A community-led protocol. Good ideas backed by real use cases get in.
Your agents are only as trustworthy as the data products they consume.
FLUID is one YAML file that describes a data product end to end — schema, build, orchestration, agentic governance, sovereignty, and semantics. Write it once; validate it, compile it, and deploy it anywhere.
The shape of a contract
fluidVersion: "0.7.5"
kind: DataProduct
id: demo.bronze.hello_world
name: Hello World
metadata:
owner:
team: data-platform
exposes:
- exposeId: hello
kind: table
contract:
schema:
- name: id
type: STRING
required: true
binding:
platform: local
format: parquet
location:
path: ./hello.parquet
→ Build it up step by step in FLUID by Example.
Where to go next
- Guide — what FLUID is, the quickstart, and the FAQ.
- Concepts — the agentic-native layer and how FLUID compares to ODCS / ODPS.
- Schema Reference — every top-level block, a cheatsheet, and the full specification.
- What's New in 0.7.5 — streaming Kafka → Iceberg sink + Confluent Cloud Tableflow.
- See the deck — the FLUID story in slides.