AT

Open source · Apache 2.0

forge-cli.

The compiler for the FLUID standard.
One contract in. Validated, planned, applied data products out — in 30 seconds.

30-second compile

Validate. Plan. Apply. Done.

forge.fish
$ pip install data-product-forge
Successfully installed data-product-forge-0.7.3

$ fluid validate contract.fluid.yaml
✓ schema  ✓ semantics  ✓ policy

$ fluid plan contract.fluid.yaml
Plan: 4 to add, 2 to change, 0 to destroy. plan-binding: 7f8a...

$ fluid apply contract.fluid.yaml
Bound. Applied. Verified. Published. Done in 28s.
The robot starves. Or worse, it makes dangerous mistakes — because it's operating on unreliable information.
— FLUID Enterprises, ch. 4 · The Tale of Two Delis

Agentic authoring

Don't write the contract.
Forge it with AI.

forge-cli ships an AI copilot. Describe your data in plain words — it interviews you, infers the schema, models the data product, and writes the governed contract for you.

forge — ai
$ fluid forge
Forge · agentic authoring run a1f3c2
AI What are we building?
A churn-risk product for the telco retention team
AI 3 sources found — customer_events is the spine. 14 fields, grain customer_id. Confirm?
:auto
modelling · inferring joins · self-healing schema
contract.fluid.yaml — 18 fields, SLO + owner set
emitted FLUID · ODCS · ODPS
validate schema · semantics · policy

It interviews you

An adaptive interview that infers before it asks — worked examples in every prompt, and :auto whenever you'd rather it just decide.

It models the product

A modeller agent infers schema, join keys and grain — productType-first. Every draft is validated and self-healed before it reaches you.

It emits open specs

One model, many outputs — the FLUID contract, ODCS and the Open Data Product Spec. It imports legacy contracts the same way.

One model — emitted in every standard

FLUID Native contract.fluid.yaml
ODCS Open Data Contract Standard
ODPS Open Data Product Spec

Run anywhere · Publish everywhere

Fluent in every cloud.

Write the contract once. Forge compiles it to native services on whichever cloud you choose — then publishes it to every catalog that matters.

Snowflake native

Databases · Schemas · Tables · Views · Tasks · Procedures · Shares · Dynamic masking & RBAC

AWS native

Glue · Redshift · Athena · S3 · IAM · EventBridge

Google Cloud native

BigQuery · Dataplex · Pub/Sub · IAM

DuckDB / local native

DuckDB-backed local builds — the identical contract on your laptop and in CI

Native to the cloud you run on today. Portable to the one you run on next.

Then publish it — to every catalog at once

Data Mesh Manager Mesh catalog, ownership & lineage
DataHub Open-source metadata & discovery
ODCS Open Data Contract Standard
FLUID Command Center Marketplace, governance & runtime
Cloud-native catalogs Glue Data Catalog · Dataplex · Snowflake Horizon

The compiler

Eleven stages. Cryptographically bound.

fluid generate ci emits a parameterised pipeline for any CI system. Apply refuses tampered plans.

  1. bundle

    Snapshot the contract + dependencies.

  2. validate

    Schema, semantics, policy gate.

  3. generate-artifacts

    DDL, dbt, IAM, policy.

  4. validate-artifacts

    Each artifact passes its native check.

  5. diff

    Resource-by-resource delta vs current state.

  6. plan

    Deterministic plan with crypto plan-binding.

    🔒 plan-binding
  7. apply

    Mode-aware execution. Refuses tampered plans.

    🔒 plan-binding
  8. policy-apply

    Native IAM bindings emitted.

  9. verify

    Quality + freshness contracts asserted.

  10. publish

    Catalog publish; consumers notified.

  11. schedule-sync

    DAG push to the chosen scheduler.

FLUID Command Center Early access

Forge them with the CLI.
Run them in the Command Center.

The runtime companion to forge-cli. Where deployed data products live, get discovered, get governed, and get observed.

Marketplace

Browse industry-tested blueprints. Install with one command. Forge yours from the same template.

Catalog

Live inventory of every data product running in your enterprise — with lineage, quality metrics, and SLAs.

Governance

Declarative agentPolicy gates, access bindings, and compliance trails — applied at runtime, not at audit time.

Analytics

Per-product, per-domain, per-agent. Know what every product costs and what every agent consumes.

You're in — we'll be in touch.

FAQ

forge-cli, answered.

What is forge-cli?

forge-cli is the open-source compiler for the FLUID standard. It takes one data-product contract and validates, plans and applies it into governed, deployed data products — in about thirty seconds.

What is a data contract?

A data contract is a declarative, version-controlled file that specifies a data product's schema, semantics, quality rules, ownership and access policy. forge-cli compiles that contract into real, native infrastructure.

Which clouds does forge-cli support?

One FLUID contract compiles natively to Snowflake, AWS, Google Cloud and DuckDB or local — and publishes to catalogs including DataHub, Data Mesh Manager and the FLUID Command Center.

Can AI write the data contract for me?

Yes. forge-cli ships an AI copilot — describe your data in plain words and it interviews you, infers the schema and writes the governed contract, emitting FLUID, ODCS and ODPS.

Is forge-cli free?

forge-cli is fully open source under the Apache 2.0 licence. There is no paid tier required to compile, validate or apply data products.

Forge your first
data product.

Install forge-cli, point it at your data, and compile in 30 seconds.