Device count
Volume tiers, counted on devices carrying the agent — not on headcount.
Core
See what is actually happening.
For a team that needs an evidenced answer to “is our data going into AI?” before it can argue for anything else.
- Browser extension capture on supported web AI tools
- Pattern detection — card numbers, national IDs, IBANs, API keys, private keys
- Log and warn enforcement
- Console dashboard, alerts and audit export
- Coverage measured per device, not assumed
No blocking, and no SIEM delivery. Alerts live in the console.
Standard
Most deploymentsEnforce it, and put it where the SOC works.
For a security team that has to stop something, map it to a framework, and have the event show up in the tool they already watch.
- Everything in Core, plus:
- Contextual detection — names, addresses and health information in free text
- Block before send, per framework and per destination
- All frameworks and your own custom patterns
- SIEM delivery — Splunk, Sentinel, Google SecOps, Elastic, CEF syslog
- SSO, RBAC and directory sync for per-department policy
The plan most organisations buy, and the one the product is designed around.
Enterprise
Cover the surfaces a browser can’t reach.
For regulated estates: desktop AI apps, internal tools calling model APIs, data residency commitments and an existing case-management workflow.
- Everything in Standard, plus:
- Local proxy mode — desktop AI apps and internal tools calling model APIs
- Data residency — US or EU, fixed at provisioning
- Dedicated tenancy rather than pooled
- ServiceNow and Jira, with disposition write-back
- Microsoft Purview labels honoured in place of inference
Proxy mode is opt-in and installs a device-trusted CA. It is a legal conversation as much as a technical one.
Not a paid feature
The privacy architecture is not an upgrade.
On-device detection, irreversibly masked snippets, mutual TLS with no plaintext fallback and the encrypted local buffer are how the product works — at every tier, including the cheapest one.
There is no configuration in which raw content is transmitted, so there is no plan in which it is. A vendor who sells you privacy as an add-on is telling you what the default does. Read the architecture.
Compare
Every difference between the plans.
| Feature | Core | Standard | Enterprise |
|---|---|---|---|
| Capture | |||
| Browser extension Chrome, Edge and Firefox, MV3 | ✓ | ✓ | ✓ |
| Local TLS proxy desktop AI apps, internal tools calling model APIs | — | — | Opt-in |
| Clipboard and window-title fallback | — | ✓ | ✓ |
| Detection | |||
| Pattern layer card numbers with Luhn, national IDs, IBANs, passports, cloud keys, private key headers | ✓ | ✓ | ✓ |
| Contextual layer names, addresses and health information in free text | — | ✓ | ✓ |
| Custom patterns project codenames, customer ID formats — encrypted under your own key | — | ✓ | ✓ |
| Sensitivity labels honoured Microsoft Purview, in place of inference | — | — | ✓ |
| Policy and enforcement | |||
| Framework mapping | PCI DSS, Internal | All eight | All eight |
| Actions available | Log, warn | Log, warn, block | Log, warn, block |
| Policy scope org-wide, by department, by group | Org-wide | Per department | Per department |
| Sanctioned-tool exemptions and allowlists | — | ✓ | ✓ |
| Console and reporting | |||
| Dashboard, alerts and case triage | ✓ | ✓ | ✓ |
| Coverage reporting whether each capture path is currently binding, not merely installed | ✓ | ✓ | ✓ |
| Audit log and export records who looked at whose data | ✓ | ✓ | ✓ |
| Retention control per framework, clamped to your contract ceiling | Fixed | Per framework | Per framework, with legal hold |
| Integrations | |||
| SIEM delivery Splunk HEC, Microsoft Sentinel, Google SecOps, Elastic, CEF over syslog | — | ✓ | ✓ |
| Pull API with cursor paging so a SIEM that was down can catch up | — | ✓ | ✓ |
| SSO and RBAC OIDC or SAML, with group-to-role mapping | — | ✓ | ✓ |
| Directory sync SCIM 2.0 — departments, and seats never enrolled | — | ✓ | ✓ |
| Notifications email, Slack, Teams | ✓ | ✓ | |
| ITSM ServiceNow and Jira, two-way, with disposition write-back | — | — | ✓ |
| Deployment and tenancy | |||
| MDM packaging Intune, Jamf, Kandji | ✓ | ✓ | ✓ |
| Data residency US or EU — immutable once provisioned | US | US or EU | US or EU |
| Tenancy model | Pooled | Pooled | Dedicated |
| Staged rollout controls ring-based, with a kill switch | ✓ | ✓ | ✓ |
| Support | |||
| Named contact and rollout assistance a person who knows your deployment, not a queue | — | — | ✓ |
Response-time commitments are set in the contract rather than published here: they differ by region, and by whether proxy mode is in play.
What drives cost
Four things move the number.
Worth knowing before a conversation, so nobody is surprised by the shape of the quote.
Proxy mode
A device-trusted CA carries a real support and engineering burden, and it is priced that way.
Residency
A dedicated regional deployment costs more to run than a pooled one.
Term length
Longer commitments price better. Nothing about the product changes.
Straight answers
Why per device rather than per user?
Because coverage is a property of a device, not a person. Somebody with a laptop and a workstation needs the agent on both, and a seat licence would quietly leave one of them uncovered while the invoice said otherwise.
Can we start on one team?
Yes, and we would recommend it. Rollout is ring-based with a kill switch, and coverage is reported per device — so a pilot tells you what your real exposure looks like before you size the rest.
What happens if we stop paying?
Agents receive a back-off response and keep buffering locally rather than discarding events, so a billing dispute does not become a coverage gap. The console goes read-only, and your data is exportable throughout.
Does a cheaper plan mean weaker privacy?
No. On-device detection, masked snippets, mutual TLS and the encrypted local buffer are how the product works at every tier. The plans differ in what Itzal can detect and do — never in what it transmits.