Compact tool call output, expandable on demand

Problem

Tool call output dominates mecatui's conversation view. Usually you don't need to see what a tool call did in that much detail, and the big blocks make the actual conversation hard to follow — reported as the single biggest reason mecatl's output feels harder to follow than other tools; what the agent is doing gets lost.

Scale makes it worse: runs regularly hit tens or 100+ tool calls depending on the model. Shrinking individual cards won't fix that — consecutive tool calls need to compact into a single line.

But they shouldn't disappear either. Early visibility is how users build trust in the agent: when switching models or trying a new process, you want to watch for it going off the rails. Once comfortable, the same output is noise. The UI has to scale across both ends.

Ideas discussed

• Compact consecutive tool calls into a single line (the core fix).
• Clickable tool cards — expand one at a time.
• Click through to a tool-call summary/detail view for real exploration.
• Not all tool calls are equal: file edits might always show the diff. Per-tool display rules, not one policy.