90 days, up to 25 seats, limits included.
A design-partner pilot puts the sensor on a real fleet and the events in your SIEM. You get the enterprise packaging and the maintainer's time. The fee is scoped in the SOW, not listed here: a number we would not take is worse than no number.
- 90
- day pilot term
- 25
- developer seats, up to
- 4h
- remote implementation support
- 2
- business days to a reply
No demo gate. No qualification call. No sales team. One maintainer reads the mail. No third SKU until a real tenant has run the MSI.
Same sensor. Different packaging.
Deploy
This page
90 days · up to 25 seats
Fleet packaging and the maintainer's time. Mail legal@agentmetry.ai. The commercial figure is in the SOW that follows, not in a cart.
What you get, and what it needs from you.
You get
Enterprise MSI
Silent per-machine install through your MDM. The ingest client ships inside the frozen binary, so a machine with no git checkout still captures.
Intune deployment pair
A Remediation that installs hooks into each signed-in user's profile rather than LocalSystem, plus the detection script, so a device fails a compliance check instead of quietly recording nothing.
Splunk TA
Search-time extraction for the heartbeat envelope: coverage per agent, hook profile, MCP schema digest, Merkle root. Saved searches ship disabled until you have a baseline.
Detections and the corpus
15 sequence rules with MITRE tags, the public benchmark they are tested against, and Splunk, Sigma and YARA-L equivalents for the heartbeat signals.
The maintainer
Up to 4 hours of remote implementation support, and a person on the other end of every message rather than a queue.
You provide
A cohort and a way to reach it
Up to 25 developer machines, and an admin who can push an MSI and a user-context script through your MDM.
An index to forward to
A Splunk HEC token and index, a Google SecOps ingestion account, or an Elastic endpoint. Nothing leaves the machine until you set one.
One technical contact
Somebody who will open the findings. A pilot where nobody triages produces a trail and no answer about whether the trail is useful.
Written feedback before the end
On detection quality and deployment friction. This is being bought as much as the fee is.
Deploy, run, review.
Deploy
MSI to the cohort, hooks into user profiles, one sink wired to your index. Success is agentmetry hooks status returning 0 on every seat and the heartbeat visible in your SIEM.
Run
Developers work normally. Findings arrive as events. You triage them and the dispositions append to the same hash chain, so the triage is itself evidence.
Review
Detection quality, false-positive rate on your traffic, deployment friction, and whether the trail answered a question you actually had. Written, so it is useful to both of us.
msiexec /i agentmetry.msi /qn agentmetry hooks install agentmetry hooks status
- 0
- every agent present here is recorded
- 1
- an agent is present and not recorded
- 2
- coverage undeterminable, install will not help
Stated here, not discovered in week three.
The MSI is not EV-signed
Deployment needs Unblock-File or your own signing. A clean-room certified build does not exist yet. If unsigned installers are a hard stop in your environment, say so first and we will not waste each other's time.
Never run against a real tenant
MSI and Intune scripts are tested against throwaway profiles. Yours would be the first real tenant. Remediations also need Intune Plan 2 or Windows Enterprise E3/E5, not base Intune.
Tamper-evident is not attributable
The hash chain proves ordering and non-modification of what was written. Per-host cryptographic identity is not shipped yet, so today a fleet trail records claims about which host sent an event rather than proof. It is an Enterprise feature and it is not built. If attributable evidence across a fleet is the reason you are here, that is worth saying in the first message.
It is not a CASB
Unmanaged ChatGPT, Microsoft 365 Copilot, and any IDE with hooks removed stay invisible. Removal is loud in the heartbeat. It is not prevented.
Covered is not capturing, for Codex
Codex trusts hooks by hash and skips untrusted ones silently. A file check cannot see that decision, so Codex seats are confirmed by events arriving, not by a coverage report.
One maintainer
No support rota and no second pair of hands. The continuity terms exist because that risk is real and should be priced rather than argued away.
Written down before anyone needs it.
Perpetual licence
The licence to the delivered build survives expiry, termination, non-renewal, and cessation of business. You keep running what was delivered.
Source deposited with you
A source archive at every delivery. You hold it from day one with no licence to use it, until a Continuity Event grants one automatically.
Apache-2.0 floor
The core is already Apache-2.0. Trails, detections and forwarders keep working with or without anyone, because nothing depends on a cloud that can be switched off.
A Continuity Event means cessation, insolvency, or incapacity. The clause lives in the pilot SOW, not in a slogan.
Evaluate it before you talk to anyone.
You do not need a pilot to evaluate this. It is Apache-2.0 and installs with pip. Run it on one machine for a week, read the trail, and decide whether the events answer a question you actually have. An evaluation that starts with a purchase order is a worse evaluation.