Skip to content
Following a session

Following a session

The timeline is where snoop starts when you open a session. It lists every prompt, reply, thinking block and tool call, one per line, and keeps up as the session writes. Each turn starts after a blank line, and a row in the same minute as the one above shows only its seconds (:42).

The timeline with prompts, tool calls and replies, and the selected reply in the detail pane
The timeline lists the session’s events, and the right pane shows the selected one in full.

Follow live

snoop --follow

With --follow, snoop opens the newest session for the current directory, or for --project DIR, and switches to a new one when Claude Code starts it. While following, the newest event is selected as it arrives and the header shows ↓. Moving up pauses following, and pressing G resumes it.

You can also open any session directly, by an id prefix or a path:

snoop 64f3ec44
snoop ~/.claude/projects/-home-me-src-api/64f3ec44-….jsonl

Read the timeline

Each tool row shows the call’s duration and status: ✓ done, ✗ error, ◌ running. Each kind of tool has its own glyph and color:

GlyphKind
▸built-in tools such as Bash, Read and Edit
✦skills
⇄MCP calls, shown as server·tool
⤷subagents

Prompt rows show the turn’s tokens, estimated cost and duration. A compaction appears as ⟳ compacted (auto) 253k → 68k.

The detail pane

The right pane shows the selected event in full: a Bash command and its output, an edit as a diff, the content of a Write, an agent’s prompt and result, a reply’s text. Press enter or tab to focus it and scroll with ↑ ↓, or scroll it from the list with J K. On narrow terminals the focused pane fills the screen.

Search and filter

Press / to search text, commands, paths and results; every word must match. Press enter to keep the search or esc to clear it. The f key cycles a filter through all, chat, tools, skills, MCP, agents and errors, and t shows thinking blocks, which are hidden by default.

Search and filters reach into subagents. A collapsed Agent call whose subagent has matches stays in the list, marked ●.

The other tabs

TabWhat it shows
1 Timelineevery event, as above
2 Toolsevery call grouped by kind, then by name, with counts, errors and average latency
3 Statsduration, tokens, cost, context, tool mix, models and a per-prompt breakdown
4 Filesevery file read or changed, with diffs. See Files and diffs

In the Tools tab, space expands a group and enter on a call shows it in the timeline.