> ## Documentation Index
> Fetch the complete documentation index at: https://docs.enfors.ai/llms.txt
> Use this file to discover all available pages before exploring further.
# The Builder lens
> The micro picture — why did this one agent do what it did?
The Builder lens is for the person who needs to know exactly **why one agent did what it
did** in a single session. It's the same data as the [Leader lens](/using/leader-lens),
seen up close.
## The Explore canvas
Everything underneath is fully explorable on the **canvas**. Start from an **agent node**
— every agent in the organization — and drill down:
See the agent, its owner, and the model it runs.
Each agent has its sessions. Pick the one you want to investigate.
A session breaks down into its decisions (e.g. 28). Walk through them in order and
see exactly what happened from the beginning.
## The logic trace
Click into a decision to open its **logic trace**: the final outcome, the action the
agent decided to execute, and the **reasoning step by step** that led there. See
[Evidence & the Logic Trace](/concepts/evidence-and-logic-trace).
## The evidence vault
Every step is backed by **evidence** — a system prompt, a git check-in, a prompt, a
document, a tool call. The evidence vault lets you reconstruct the entire decision from
its logic and the evidence behind it.
## Session quality & speculation
Enfors scores the quality of every session — not just standard metrics like confidence
and routing, but advanced ones like the **speculation ratio**: the share of decisions
*not* backed by evidence. It's how you catch errors before a customer does. See
[Insights & Drift Detection](/concepts/insights-and-drift).
## Do it programmatically
The same drill-down is available via the [MCP server](/mcp/overview):
```text theme={null}
list_agents → list_agent_sessions → list_session_decisions → get_decision
```
For relationship traversal (evidence, findings, steps), use
[`query_graph`](/mcp/tool-reference#query_graph).