← c12c12 / wf1 - Inventory Deep Dive / Inventory / Agents / Detail
Qpoint
QI
Inventory /Agents/claude-code@1.2.0

Agent Detail

An agent detail page should explain one software lane end to end: who runs it, where it executes, what stack it depends on, and how its behavior is changing.

The page should help a reviewer see the agent as a repeated organizational pattern, not just a row with a version and session total.

Use This Page When Asking

What does this software lane mean in practice, and where is its behavior drifting from the expected baseline?

Open c6 raw detail

Top Read

Claude Code is the dominant coding lane across the observed org footprint.

It appears across 12 users and 31 sessions, usually with a stable stack. The review trigger is not ordinary usage itself, but one unknown operator case plus a pending version change that could shift the baseline quickly.

Org reach

31 sessions

Highest-volume agent lane

Broad use

12 users

Broadest human reach in scope

Open anomaly

1 unknown operator

Needs comparison against the stable lane

Drift watch

new build pending

Stable since Mar 15, but baseline may move soon

Why This Detail Matters

This detail should tell the reviewer what this agent is, how far it reaches across the org, and what changed recently enough to justify deeper review.

Agent Identity And Baseline

Establish the software family and its normal operating shape first, then judge whether current usage still fits that baseline.

Agent

Claude Code

claude-code@1.2.0

Version

v1.2.0

SDK Language

js / node

SDK Runtime

node

SDK Version

0.38.0

Agent ID

claude-code@1.2.0

What This Detail Should Help Decide

The page should leave the reviewer with a concrete next move, not just more context.

Inspect the unknown operator case

Check whether the unresolved user or endpoint still fits the normal Claude Code lane, or whether it opens a genuinely new trust question.

Open c6 raw detail

Compare the pending build against the stable baseline

Use the raw version inventory to separate ordinary upgrade motion from behavior drift that changes risk meaning.

Open c6 raw list

Validate the lane through top attached sessions

Review the few recent sessions and user rollups that best confirm whether this is normal expansion or surprising new reach.

Open session detail

Compared With Other Agent Families

This page should help the reviewer judge whether Claude Code is merely common or uniquely consequential in the current inventory.

Use comparison to judge consequence, not just popularity.

Session lead

31 vs 8

Nearly 4x the next-highest family

User lead

12 vs 4

Broadest human reach in the dataset

Open anomaly

1 unknown operator

Small in count, high in interpretive value

Shared dependency

filesystem MCP

Top attached tool surface across the lane

Observed Behavior

This section should explain what the agent actually does in practice: which surfaces it uses, how much leverage it has, and where ordinary work turns into stronger review interest.

Behavior is often the fastest way to understand leverage.

Tool Surface

Mostly code-oriented local actions

The normal lane is file reads, file edits, and shell execution. That makes the agent powerful even before any external traffic enters the story.

MCP Reach

Filesystem MCP is the dominant attached server

Most observed usage flows through a narrow, understandable tool surface. That is good for baseline clarity, but it also means sensitive files are close by whenever the lane is active.

Delegation Pattern

Occasional fan-out into subagent work

Subagent launches are not the default read, but they matter because they multiply execution paths and make one agent lane feel more like a small workflow system.

Boundary Behavior

Usually local-first, sometimes outbound after context build

Most activity stays on host and repo surfaces. The more consequential moments happen when sensitive context is already in scope and the lane then reaches outward to models or secondary APIs.

Cost And Usage Shape

Agent economics matter most as an aggregate behavior signal. The question is not just whether this lane spends money, but whether its cost profile matches the kind of work and reach the agent is supposed to have.

Use economics to understand work shape, not just spend.

Aggregate Posture

Broad reach matters more than raw spend alone

Claude Code appears across many ordinary coding sessions, so its total cost should be read alongside breadth of use. A widely used agent can be operationally important even when it is not the single most expensive lane.

Usage Shape

Mostly steady Sonnet-scale work, with a few heavier runs

The normal pattern is many repeat coding sessions on the default model lane. The notable exceptions are the fewer attached sessions where cost or model choice rises sharply enough to test whether the baseline is shifting.

Outlier Read

One higher-cost attached run is more useful than average-spend math

The raw agent evidence already shows a heavier Opus-backed session. That kind of outlier is usually a better review trigger than a blended average because it tells you when the lane may be doing more than its normal work.

Agent-Centered Topology

An agent map should connect the software to its users, hosts, dependencies, and change risk.

Useful per agent: who runs it, where it appears, which models and credentials it prefers, what capabilities it activates, and whether a new version or behavior spike stands out.

Agent

Claude Code v1.2.0

Dominant org-wide coding agent with the broadest footprint across users, endpoints, files, and MCP servers.

31 sessions · 12 users

Primary SDK

js / node

Common model

claude-sonnet-4-6

Common key

sk-ant-...A3F2

Version posture

stable · seen since Mar 15

Operators

Who this agent belongs to in practice.

3 nodes
alice@company.com6 sessions
carol@company.com5 sessions
unknown (10.0.1.44)1 session

Runtime Surfaces

Where the software has executed.

3 nodes
alice-mbp.localMacOS arm64
dave-mbp.localMacOS arm64
10.0.1.44unknown endpoint

Dependencies

What the agent tends to invoke.

3 nodes
claude-sonnet-4-6preferred model
filesystem MCPhigh-use server
read_file / bashtop capabilities

Watch For

Signals that change the meaning of usage.

3 nodes
new version arrivalupgrade event
error spiketool failures
unknown operatortriage path

Representative Event Chain

The page should show how the item participates in a readable sequence, not just a pile of supporting rows.

5 steps
  1. 1

    Baseline

    Claude Code is the shared coding baseline

    This family is no longer just one app install. It is the default execution lane across the broadest set of observed users and developer hosts.

    31 sessions12 usersalice-mbp.local
  2. 2

    Reach

    Its reach spans people, hosts, and sensitive work surfaces

    Because the lane appears in many sessions, one change in its behavior can affect how a large slice of the inventory should be interpreted.

    dave-mbp.localauth-adjacent filesshared org key
  3. 3

    Dependencies

    The lane has a recognizable operating shape

    The common model, key pattern, and top tool surface give reviewers a baseline for what ordinary Claude Code usage looks like.

    claude-sonnet-4-6filesystem MCPread_file / bash
  4. 4

    Drift

    The main review question is whether current usage still fits that baseline

    A pending new build and one weakly identified operator are enough to move this page from background inventory into an active review surface.

    new version pendingunknown operator10.0.1.44
  5. 5

    Verify

    The next step is to inspect the few attached sessions that best test the baseline

    Once the lane is framed, recent-session rows and user rollups become confirmation tools for whether this is ordinary expansion or a real change in reach.

    recent sessionsuser rollupc6 raw detail

c6 Evidence Tables

These are verification surfaces after the story is understood, not the first read.

Recent Sessions

c6 shows where the agent appears in recent live and ended work.

UserEndpointModelCostStatus
alice@company.comalice-mbp.localclaude-sonnet-4-6$0.43active
carol@company.comcarol-mbp.localclaude-sonnet-4-6$0.28active
bob@company.combob-thinkpad.localclaude-opus-4-6$4.10ended

Users Running This Agent

The raw page also rolls the lane back up by person.

UserSessionsTotal costLast seen
alice@company.com6$2.14today
carol@company.com5$1.82today
eve@company.com4$1.40today

Return To Category

Compare this representative detail against the full category framing and drill targets.

Open Agents overview

Open c6 Raw Detail

Use the original detail page for exact raw evidence after the c12 story framing.

Open c6 raw detail

Open c6 Raw List

Compare this item against the broader c6 entity ledger once the representative story is clear.

Open c6 raw list