Watchlight AI
Back to Platform OverviewTier 4: Lineage

Trace How Every Action Happened

When your CISO asks "how did that agent get access to our CRM?", Execution Lineage gives you the full chain in seconds: from the human who authorized it, through every delegation hop and policy decision, to the downstream resource.

The Reconstructability Gap

Authorization tells you if an action was permitted. Lineage tells you how it came to be.

The Missing Middle

You see the first request and the final API call. Everything in between is invisible. Delegation hops, policy decisions, and credential use are disconnected events.

Reconstruction Takes Hours

When an incident occurs, teams spend hours correlating timestamps across systems to reconstruct what happened. By the time the chain is assembled, the damage is done.

Compliance Cannot Prove Authorization

Auditors ask "who authorized this agent to access that system?" The answer requires forensic archaeology across logs, policy records, and delegation grants.

How It Works

A tree-structured audit model that connects every delegation hop, policy decision, and resource access into a single reconstructable chain.

Tree-Structured Audit Model

Every agent action is a node in a directed graph. Parent-child edges represent delegation. Each node carries execution identity, agent identity, declared intent, policy decision, and timestamp.

Cross-Boundary Propagation

Lineage context propagates across service boundaries via lightweight HTTP headers. Services that do not understand lineage ignore the headers. No modifications required.

Policy-Aware Context

Lineage attributes are available to the policy engine during evaluation. Policies can reference delegation depth, chain validity, intent category, and trust state.

Persistent Execution Graph

Executions are stored as a tree with parent-child references. Recursive queries reconstruct the full chain from any node. One execution ID returns the complete lineage.

Anomaly Detection

The lineage graph automatically flags suspicious patterns: broken chains, denied-after-allowed-parent, deep delegation, unusual fan-out, and orphan executions.

Tamper-Proof by Design

Lineage records are append-only and immutable once written. Server-authoritative fields are never trusted from client requests. Anti-spoofing controls prevent chain forgery.

NewIn Early Access

The Agent Execution Graph

A queryable graph of every governed agent action, built from event streams as they flow through the runtime control plane.

Real-Time Visibility

See every agent action as it happens, with full governance context attached.

Forensic Reconstruction

Query by any execution ID to get the full chain back: every hop, every decision, every resource.

Anomaly Alerting

Suspicious patterns surface automatically: broken chains, deep delegation, unusual fan-out.

Cross-Service Correlation

Events from proxy, policy engine, and agent frameworks correlated by a single execution ID.

Compliance Evidence

Every node carries identity, intent, authority, policy version, and timestamp.

Event-Driven Architecture

Built on NATS JetStream with durable consumers, append-only events, and replay support.

Three Questions It Answers

For CISOs: "Who authorized that agent to access our CRM?"

The lineage chain shows the full path: Sarah Chen (analyst, authenticated via Okta SSO) delegated to the orchestrator, which sub-delegated a CRM query to the senior-researcher agent, authorized by the research-access policy under a DataAnalysis intent within the Q4 Research goal. Every hop audited. Full chain reconstructable from a single execution ID.

For Compliance: "Can we prove this action was authorized?"

Yes. The execution record links to the delegation grant (who delegated, what scope, when it expires), the policy evaluation (which policies were checked, which one decided), the intent declaration (why the agent claimed it needed access), and the goal (under what business objective). All persisted, all queryable, all immutable.

For Platform Teams: "How do I add lineage to my agents?"

Start with zero changes. The governance proxy auto-generates execution IDs for every request. For delegation tracking, add one header to sub-agent calls. For intent tracking, add one more. All optional. All additive. All backward compatible.

Automatic Anomaly Detection

The lineage graph flags suspicious patterns in real time, giving security teams immediate visibility.

PatternWhat It MeansSeverity
Broken chainParent execution referenced but not foundWarning
Denied after allowed parentSub-agent denied when parent was allowedInfo
Deep delegationChain depth exceeds thresholdWarning
Multiple sibling delegationsUnusual fan-out from single parentWarning
Orphan executionExecution with no associated decisionsInfo

Key Outcomes

Reconstruct any delegation chain from a single execution ID
Answer "who authorized this?" in seconds, not hours
Policy decisions that reference the full delegation context
Automatic anomaly detection across the lineage graph
Compliance-grade evidence of every authorization chain
Zero-change adoption via proxy-generated execution IDs

Grounded in Governance Principles

Every capability traces back to the 12 non-negotiable principles for Agent Runtime Governance.

Every Action Explainable. Every Chain Reconstructable.

Execution Lineage is the difference between "the AI accessed your CRM" and a complete, auditable chain of who authorized it, how it got there, and why.

Request a Demo

We value your privacy

We use cookies to enhance your browsing experience, analyze site traffic, and personalize content. You can choose to accept all cookies or customize your preferences. Learn more