Skip to content
Agent Archaeology

Foundations

Learn

Agent Archaeology studies what AI agents saw, decided, and did. Agents now operate across editors, terminals, browsers, filesystems, APIs, and internal tools. When an agentic event matters, teams need a practical way to reconstruct what happened.

What the field studies

An agentic event may include user intent, hidden instructions, retrieved context, workspace state, tool calls, model outputs, logs, and downstream side effects. Agent Archaeology connects those artifacts into a defensible timeline.

Why it matters now

A single agent task can move from prompt to model response to shell command to file edit to API call. Traditional logs capture fragments. The evidence is distributed across session stores, IDE extensions, terminal history, MCP server logs, source control diffs, and audit trails.

Investigation posture

Treat agent records as evidence, not as truth by default. Preserve original artifacts, document transformations, and mark uncertainty clearly. Separate what was observed from what was inferred.