Handeware Scanner · handeware-scan

Find out what your estate actually is.

Before anyone can modernize a workflow estate, somebody has to say what is in it. Scanner reads the exports you already have — modeled processes, batch schedules, and the operational exhaust nobody ever modeled — and returns an inventory, a classification of every task, and an agent-readiness score per process. It is free, it runs offline, and it needs no account.

Price: Free
Dependencies: None — CPython only
Network: Never
Install: One command

Try it on a synthetic bank before you point it at yours

pip install handeware-scanner
handeware-scan --demo

That writes a full report for "Demo Bank" — a card-dispute process, a KYC review, a wire-exception repair, an end-of-day Control-M batch, an AutoSys payment release chain, plus mailbox and file-share exhaust — then tells you exactly how to point it at your own exports. Nothing leaves the machine, so it is safe to run inside a sandbox while you are still deciding whether to trust us.

Read the report it produces →

What it reads

Tier	Source	What we take from it
Modeled	BPMN 2.0, DMN	Processes, tasks, lanes, gateways, boundary events, decision tables
Modeled	IBM BAW `.twx`, Pega exports	Process definitions extracted from the archive; Pega assets detected
Batch	Control-M XML, AutoSys JIL, crontab	Job chains as workflows — conditions are sequence flows in disguise; human-gate keywords surface choke points
Discovered	Case-system event logs (CSV)	Mined process models: variants, rework ratio, handoffs, cycle time
Signals	File-share listings, mailbox exports	Recurring artifact cadence and correspondence clusters — workflows that exist without ever being modeled

Irrelevant files are skipped silently. Formats are auto-detected — you do not tell it what it is looking at, and it never asks you to normalize your estate first.

How it classifies

Every task lands in exactly one of five buckets, and each carries a one-line rationale so a reviewer can argue with it:

Classification	Meaning
`KEEP_DETERMINISTIC`	Already automated. Leave it as code or rules — this is where the replay proof applies.
`AGENT_CANDIDATE`	Human judgment an agent can take on.
`AGENT_WITH_HITL`	Agentable, but a human must stay in the loop.
`NEEDS_DISCOVERY`	The signal is too weak to decide. Assess before committing.
`HUMAN_ONLY`	Physical-world or mandated human step.

Processes score 5–95 for agent-readiness, with the score drivers listed rather than hidden behind a number. The gap analysis then tells you the uncomfortable part: how many workflows your modeled estate does not contain at all.

Why free

Because an assessment nobody can run is worth nothing, and because the product we sell is what comes after the assessment. Scanner is deliberately the most spreadable thing we make: no license server, no seat count, no phone-home, no expiry. An engineer can run it on Friday afternoon without asking anyone, and the report is designed to be forwarded upward without us in the room.

The next question it plants

Every report ends by naming the deterministic tasks in your batch chains and the one command that would prove them — handeware-prove. That is the whole funnel, printed in the artifact rather than tracked in a beacon.

Run it in a container

docker run --rm --network none \
    -v "$PWD/exports:/work" handeware/scanner .

--network none is not a hardening tip we suggest; it is the reference deployment, and the full test suite passes under it. See the trust page for the rest of the posture.