Use cases

What you can build.

Five patterns MachineCraft is built for — each on the same two engines, the same audit trail, and the same deploy-anywhere foundation. These are what you can build, not customer case studies: we’re pre-launch and won’t pretend otherwise.

Join the waitlist Browse templates

What you can build

Five ways to put it to work.

Each pattern names the engine it runs on, who it’s for, the steps to build it, and the outcome — with a real starter template to begin from. Beta capabilities are labeled beta, every time.

UC-01 / DOCUMENTS

Workflow Engine

Document processing with a compliance audit trail

Turn stacks of contracts, filings, or reports into grounded answers, while the platform logs every parse, model call, and credential access behind them.

How it works

Ingest a document, parse it, and ground a model in its contents — built visually from 100+ components, no glue code.
Every action lands in an action-level audit trail (SOC 2 Type I aligned), so you can show exactly what was read and what was produced.
Add a human approval gate (beta) where a regulation requires a reviewer to sign off before a result is used.
Run the whole pipeline inside your own perimeter — cloud, on-prem, or air-gapped — with no outbound calls.

Outcome

Answers you can trace back to the source, with the evidence an auditor will ask for.

For regulated enterprises

TemplatesDocument Q&A Vector Store RAG

UC-02 / SUPPORT

Workflow Engine Agent Engine· beta

Customer support that escalates to a human

BETA

Handle common requests with a grounded assistant, and put a reviewer in the path of anything that needs judgment — a refund, a policy exception, a regulated disclosure.

How it works

Ground replies in your own help content and the conversation so far, so answers stay on-policy.
Require human approval (beta) on the sensitive steps — a reviewer is notified to approve or reject before a result reaches the customer.
Every escalation, decision, and reviewer identity lands in the audit trail.
Keep customer data inside your environment — deploy the same flow wherever it has to run.

Outcome

Fast answers where it’s safe to automate, a human in the loop where it isn’t — and a record of which was which.

For regulated enterprises

TemplatesMemory Chatbot Document Q&A

UC-03 / DATA

Workflow Engine

Data extraction and transformation pipelines

Build repeatable pipelines on the Workflow Engine that read your files, parse and clean what’s inside, and ground a model in the result — the structured-automation half of the platform.

How it works

Compose ingestion, parsing, and model steps visually — 100+ components across 22 model integrations.
Schedule pipelines to run on a cron, or trigger them from your own systems.
Drop into code wherever a transformation needs it — the visual flow and custom code share one surface.
Promote the same pipeline from a test environment to production unchanged.

Outcome

A repeatable pipeline that turns raw documents into model-ready data — built once, run anywhere.

For research teams

TemplatesVector Store RAG Document Q&A

UC-04 / RESEARCH

Agent Engine· beta

Research agents that reason in steps

BETA

The Agent Engine (beta) plans, calls tools, and loops toward an answer, while you keep oversight over the consequential calls.

How it works

Chain specialist steps — gather, analyze, synthesize — and let the agent decide the path through them.
Inspect the agent’s decisions and intermediate state as it runs.
Require approval (beta) before the agent takes a consequential action.
It’s in active beta — powerful, and still maturing; we label it that way everywhere.

Outcome

A head start on open-ended research, with a human in control of the calls that matter.

For research teams

TemplatesSequential AgentBETA

UC-05 / DEPLOY

Workflow Engine Agent Engine· beta

One solution, deployed across every client environment

BETA

For teams that deliver AI to many clients at once: design a workflow or agent once, then ship the same portable artifact into each client’s cloud, on-prem, or air-gapped environment.

How it works

Workflows are portable artifacts — the same definition deploys to any environment, with env-var-only differences.
No phone-home runtime, so a client’s air-gapped network is never a blocker.
Carry the same governance into every deployment — approval gates (beta) and the audit trail travel with the workflow.
Standardize your team on one platform across clients instead of rebuilding per environment.

Outcome

Design centrally, deploy everywhere, and meet each client’s compliance bar with the same solution.

For it & service providers

TemplatesBasic Prompting Sequential AgentBETA

Keep exploring

Where to go from here.

Start from a template

Every use case above starts from a real, working starter flow. Browse the gallery and open one.

Browse templates

Find your team

Regulated enterprises, IT & service providers, research teams — see the platform through your mandate.

Explore solutions

See how we compare

Honest, side-by-side comparisons against the tools you’re weighing MachineCraft against.

Compare tools

How to read this

What you can build, not who’s done it.

We’re in private beta with no public customers, so everything here is framed as a capability you can build — never a case study we can’t back up. The Agent Engine and human-in-the-loop approval gates are in beta and labeled that way wherever they appear: the approve/reject flow is functional and demoable today, while durable, cross-restart execution is still on the roadmap — so an in-progress approval won’t survive a restart yet, and we don’t claim it does. Compliance work is described as “aligned,” never “certified.”

SHIPPEDBETA

The Workflow Engine and audit trail are shipped; the Agent Engine and approval gates are beta. See the roadmap for what’s next and the changelog for what’s already landed.

Build your first one during the beta.

MachineCraft is free during private beta. Join the waitlist, or tell us about the workflow you have in mind.

Join the waitlist Talk to our team