mesh

v2.4.1

open-source · apache 2.0 · v2.4.1

Data ownershipat the domain.Pipelines that disappear.

Mesh hands data ownership back to the teams who understand it. Federated governance, self-serve infrastructure, zero central bottlenecks.

Star on GitHub Read the Docs

$ npm install @mesh-framework/core

scroll

live metrics

Traction before explanation.

# numbers pulled from GitHub API and npm registry

+5.1%

GitHub Stars

+1.2k this month

+18.4%

Monthly Downloads

via npm & PyPI

+9.7%

Production Deployments

tracked last 30 days

+12 this week

Contributors

across 28 countries

architecture

Four pillars.
One coherent system.

Each concept maps directly to Zhamak Dehghani's data mesh principles — with production-ready tooling, not just theory.

Data Products

Ship data like you ship software.

Each domain publishes a versioned, SLA-backed data product. Consumers subscribe — not scrape. Ownership is explicit. Deprecation is graceful.

v3.0

v3.1

v3.2

order_events

SLA ✓ 99.9%

analytics

ml-team

"code-comment"># domain: ordersproduct:  "code-property">name: order_events  "code-property">owner: orders-team  "code-property">version: "3.1.0"  "code-property">sla:    "code-property">freshness: 5m    "code-property">uptime: 99.9%

Federated Governance

Global standards. Local enforcement.

A global policy plane defines interoperability contracts. Each domain enforces them autonomously — no central team required.

Global Policy Plane

orders

payments

users

↑ enforces locally

mesh policy apply \  --scope global \  --rule pii-masking \  --enforce auto

Self-Serve Platform

Provision in minutes, not sprints.

Domain teams get a full data stack — storage, compute, observability — via a single CLI. No tickets, no waiting.

$ mesh domain init

storage ✓

compute ✓

observe ✓

Ready in 2m 14s

mesh domain init payments \  --stack postgres+dbt \  --observe on

Domain Ownership

The team that builds it, owns it.

Eliminate the central data team bottleneck. Domain engineers own pipelines, quality SLAs, and the data contract — end to end.

payments-team

pipeline: stripe_events

quality: 99.97%

freshness: 1m 42s

from mesh import Domaindomain = Domain("payments")domain.own(  pipeline="stripe_events",  sla={"freshness": "2m"})

code first

Every concept
ships as code.

YAML, CLI, SDK, or OPA policy — pick the interface that fits your team's workflow.

Define a data product in one file. Mesh handles the rest.

data-product.yaml

# Define your data productapiVersion: mesh.io/v1kind: DataProduct metadata:  name: order_events  domain: orders  owner: platform@orders.team  version: "3.1.0" spec:  source:    type: kafka    topic: orders.v3.events   output:    format: delta    partition_by: [event_date, region]   sla:    freshness: 5m    completeness: 99.95%    uptime: 99.9%   governance:    pii_fields: [customer_email, billing_address]    classification: internal    retention: 90d

in production

Teams shipping with Mesh.

Engineering leads at Series C+ companies on what changed.

We cut our data team's pipeline ticket backlog by 80% in two months. Domain teams now ship data products the same way they ship services — independently.

80%fewer tickets

Marcus Okonkwo

Staff Platform Engineer · Ramp

Mesh is the first OSS project that actually operationalizes the data mesh paper. Not a whiteboard exercise — real tooling with real contracts.

6 weeksto production

Priya Venkataraman

Principal Data Architect · Figma

Our CTO read the Dehghani book and asked us to implement it. Mesh turned that conversation from a 12-month roadmap into a 6-week rollout.

12→6wkimplementation

Tobias Lindqvist

VP Engineering · Klarna

trusted atRampFigmaKlarnaVercelLinearRetoolNotion

deploy is green

Your data lake has
been holding you back.

24,800 engineers already pushed the first commit. The repo is open, the docs are honest, and there's no enterprise edition to upsell you into.