Claude Code: Live Running-Agent Count in /agents and Bridge Session Git Context

Live Subagent Count in /agents

Claude Code v2.1.97 introduces a ● N running indicator displayed inline in the /agents panel, adjacent to each agent type that currently has live subagent instances. Previously, there was no native way to see at a glance how many background agents were executing — developers had to infer this from tool call output or rely on third-party plugins.

The indicator updates in real time, making it easy to track the parallel workload at a glance during complex orchestration tasks. This is particularly useful when running custom subagent types (Explore, Plan, Task) in parallel, since a user can open /agents and immediately confirm which types are active and how many instances each has spawned.

This feature directly addresses a long-standing community request tracked in the Claude Code GitHub issues — the feature request asked for a persistent active subagent counter visible from the CLI status line. The v2.1.97 implementation surfaces it within the /agents panel rather than the status line, but the core visibility gap is resolved.

Bridge Sessions Now Surface Local Git Context

For developers using Bridge sessions — Claude Code's mechanism for attaching a local terminal session to a remote claude.ai session card — v2.1.97 now displays the local git repository name, current branch, and working directory on the session card. Previously, Bridge session cards showed only a generic identifier, making it difficult to distinguish between multiple simultaneous Bridge sessions from different projects or branches.

With this change, a developer running three Bridge sessions across three feature branches can identify each session at a glance from the claude.ai dashboard without needing to inspect them individually.

Context and Significance

Both improvements target the multi-agent and distributed-work patterns that have become increasingly central to how Claude Code is used. As agentic workflows grow more complex — with orchestrator agents spawning parallel subagents, and developers bridging local environments into cloud sessions — visibility and session legibility become critical. These additions are small in implementation surface but high in practical impact for teams running Claude Code at scale.