OctoMesh BY MESHMAKERS.iO
The AI-ready data platform

There is no AI
without data.

OctoMesh makes your data AI-ready — machine signals, meters and documents alike, from raw value to digital twin to the agent that understands it. Model, refine, serve: one platform instead of a Frankenstack.

Frankenstack /ˈfɹæŋkənstæk/ — Frankenstein + stack. A data architecture stitched together from tools that were never meant to fit. It works. Until it doesn't.

In production with partners and customers
COPA-DATA Microsoft Partner solbytech FH Salzburg iSPro-NG IoTechnik
30+products live
50+tenants in production
10M+pipeline executions per year — sensors to documents
13open-source agent skills on GitHub
180+MCP tools for AI agents
One semantic model

Model once. Serve everywhere.

Your domain knowledge becomes a typed, versioned Construction Kit — a digital twin of your machines, sites, documents and processes. That one model powers every consumer at the same time:

Dashboards

Meshboards visualize live and historical data straight from the model — no BI glue layer.

Applications

Custom web apps ship as blueprints and deploy to any tenant with one command.

GraphQL API

Every type, association and time series is queryable the moment you model it.

AI agents

Agents read the same governed model via MCP — with real identity, roles and audit.

OctoMesh for AI

Other platforms bolt AI on.
OctoMesh gives agents a governed home.

Running in production today — not a roadmap slide.

Agent Access

MCP

Claude & Co. operate OctoMesh directly — explore the data model, query instances, administer tenants. 180+ tools, secured by OAuth, roles and delegation.

AI Pipelines

ETL + LLM

LLM steps as pipeline nodes: understand documents, classify data, match records — under a traceable identity instead of a shadow admin.

Semantic Model

CONSTRUCTION KITS

Your domain knowledge as a typed, versioned model — the context agents need to understand data, not just read it.

Governed Autonomy

IDENTITY + AUDIT

Every agent action runs under a real identity with fine-grained data permissions and a full audit trail — AI without losing control.

How it works

From sensor to agent — in four steps.

01 · COLLECT

Edge & adapters

Connectors for OPC UA, MQTT, files and REST APIs — managed by a Kubernetes operator, scaling to zero when idle.

02 · MODEL

Construction Kits

Types, associations and enums as a versioned semantic model — your digital twin, designed in Refinery Studio.

03 · REFINE

Pipelines & archives

Declarative data flows transform, match and enrich; time-series archives keep every measurement queryable.

04 · SERVE

APIs, apps & agents

GraphQL, dashboards, custom apps and MCP for AI agents — all speaking the same governed model.

Developer Hub

Built for people
who build.

Guides for users, operators and developers, a full API reference and a 5-minute quick start. Model with the Studio, automate with octo-cli, ship as blueprints — and let your AI assistant do the routine work via MCP.

.NETGraphQLKubernetesAngularMQTTOPC UA
Proof, not promises

30+ products run on OctoMesh — here are three.

All solutions →
BUILT ON OCTOMESH

OctoMesh Energy

PV, storage and loads connected to one smart energy system — forecasts, day-ahead prices, automatic optimization, peak shaving.

energy.meshmakers.io →
BUILT ON OCTOMESH

BelegCockpit

AI-assisted accounting: documents understood, transactions matched, bank feeds reconciled — receipts to reports, automatically.

accounting.meshmakers.io →
BUILT ON OCTOMESH

Cyber Resilience

CRA compliance for product makers — structured evidence, assessments and reporting on a governed data platform.

cyber.meshmakers.io →
vibeclean — a meshmakers service

Your prototype runs.
Nobody dares to deploy it.

Cursor delivered. Claude delivered. What's missing: someone who takes AI-generated code to production grade — before your first real user finds the gaps. Audit, refactoring, hardening: by the team that runs OctoMesh in production.

Ready to refine your data?

Start in five minutes, or let us walk you through the platform on your own use case.