Skip to content
MoorAI
Get MoorAI
MoorAIKnowledge & evidence
On this page
Event Flow— Source → Actor → Destination, content-freeLineage, not surveillanceEvery edge is redacted metadataContent-free by construction
// moorai · event flow

See the flow, not the content.

Data lineage for AI agents — without the surveillance. Security tools show you where data went by reading it. MoorAI draws the same picture from content-free signals alone: which app a coding agent read from, which actor drove it, where it was headed, the data class, and the risk. The fields are category, risk level, a one-way actor hash, tool, and stage. No filenames, no prompts.

MoorAI — Security Dashboard OverviewTimelineEvent FlowPolicyCompliance

Event Flow — Source → Actor → Destination, content-free

Credential-class data read via hook:Read → headed to an external coding agent · blocked
Local → External agent
SOURCE ACTOR DESTINATION hook:Read Credential · 42 events · 6 risky hook:Bash Secret · 18 events mcp__slack Source code · 11 events claude -p PII · 27 events UBA Actor 7f3a 60 events · 6 risky ⛔ BLOCK ENFORCED Actor b1c9 38 events · 2 risky external model API egress · 31 events agent output output · 22 events agent context file · 16 events blocked 6 events · policy: block
allowed elevated blocked / critical Content-free — nodes are apps, actor hashes, and stages; never filenames or prompt content.

Lineage, not surveillance

Cyberhaven-style lineage traces data from origin to exfiltration, but it reads the data to do it. MoorAI traces the same path with nodes, direction, and counts, and doesn't need the filename.

Every edge is redacted metadata

Source = the tool/MCP the agent read through. Actor = a short fingerprint of the operator, used to group one person's events together. It is not an anonymizer: the username and hostname are recorded alongside it in clear. Destination = the egress stage. Data class + risk color it. That is the whole graph, and none of it is content.

Content-free by construction

You don't have to take "we don't store your prompts" on faith. The agent is open source (MIT), and the telemetry is redacted before it leaves the machine, so there is nothing stored to leak.

MoorAI overview → On-device AI DLP → OWASP LLM Top 10 →

The dashboard view shown is representative; nodes, actors, and counts are illustrative. Every element derives from content-free alert fields (category, risk level, actor hash, tool, stage) that MoorAI already emits — no filename, prompt, or file content is stored or displayed. Cyberhaven is a trademark of its respective owner; this page is not affiliated with or endorsed by it.