Group repeated Agent tool calls in the panel

[anilzeybek]

What are you proposing?

Add a way for the Agent panel to group or stack consecutive similar tool calls, such as multiple Read file, List directory, or search/grep calls.

For example, instead of showing many separate rows like:

• Read file crates/foo/src/a.rs
• Read file crates/foo/src/b.rs
• Read file crates/foo/src/c.rs

Zed could show a compact collapsed group such as:

• Read 3 files

Clicking the group would expand it to show the individual tool calls and their details.

Why does this matter?

Agent sessions often involve many small file-reading or directory-listing operations before the agent gives a useful answer. Even when the actual tool output is collapsed, the repeated tool-call headers take up a lot of vertical space and make it harder to scan or navigate the conversation.

This is especially noticeable during codebase exploration, where the useful human-facing messages can get separated by long runs of repetitive tool activity.

Are there any examples or context?

The existing agent.expand_terminal_card setting helps a lot for terminal output. This request is for a similar reduction in visual noise, but at the tool-call-list level rather than only hiding a single tool call's output.

Possible approach

Group consecutive completed tool calls by kind within the same agent turn/session, while keeping failed, pending, permission-gated, or user-actionable calls visible individually.

A collapsed group could show the tool kind, count, and maybe a short summary such as:

• Read 8 files
• Searched 4 times
• Listed 3 directories

Expanding the group would reveal the current individual rows unchanged.