RUTILE

Discover. Govern. Prove it.

Identify AI agents within a second, decide by behavior within 100ms, and seal every decision.

endpointDev-PC-02
Identified 0/3

Launch a tool. It is detected in about a second.

Processes

  • No AI tools running

Evidence layers

Select a process to see its evidence

  1. Binary signaturePending
  2. Process contextPending
  3. Network fingerprintPending

An identity within a second of launch

The endpoint agent detects AI tools instantly from OS kernel signals. Binary signature, process context and network fingerprint combine into a persistent identity (DID).

  • Automatic detection, no opt-in
  • Identity survives tool updates
  • Signer swaps are flagged instantly

Policy on behavior, not vendors

One policy applies the same rule to every AI tool. Decisions and enforcement finish on the endpoint within 100ms.

Policy serverendpointbundle

Policy server connected. The bundle is deployed on the endpoint.

Last decision0 · 100ms
Waiting

Click a cell and the endpoint decides that behavior for that tool.

  • Sensitive file read

    scenario-1-secret-exfiltration

  • External git push

    scenario-2-external-git-push

  • Unverified package install

    scenario-3-unsigned-package-install

  • Unapproved model call

    scenario-4-unapproved-model

  • Unregistered MCP connection

    scenario-5-unregistered-mcp

One policy applies to all three tools

Every decision, sealed immutably

Who, when and what: every policy decision and its evidence is sealed into the log. The log is insert-only, so nothing can be erased or edited. It meets the audit-trail mandates of ISMS-P and financial regulators out of the box.

  • Matched evidence and approver recorded
  • Checks modeled on the EU AI Act and OWASP LLM Top 10
  • Searchable audit console
Audit loginsert-only
Chain verified6 / 6 events

Flip block 3's verdict to 'allowed'. The previous-hash of every later block stops matching and the chain breaks.

Evidence

Who
claude.exe · dev01 · Dev-PC-02
When
09:14:07
What
File read customers.csv BLOCKED
Matched policy
scenario-1-secret-exfiltration
Evidence
File read: c:\test\data\customers.csv (SENSITIVE)
Approver
None
Hash
72d71e6a0cc81458
Previous hash
64879517a5de67ad

Thirty minutes is enough

We run a GABA Governance demo tailored to the AI tools your teams use.