Claude Code Vitals
Real-time session status detection, multi-session management, and compact tracking for Claude Code.
English | 日本語
Install
Search Claude Code Vitals in VS Code Extensions, or:
ext install barrel1054.claude-code-vitals
Requirements
Features
Session Cards
Each session is displayed as a card with real-time status, progress bar, and metadata.
| Element |
Description |
| Session name |
AI-generated title, first user input, or session ID |
| Elapsed time |
Relative time since last update (<1m / Xm / Xh) |
| Status icon |
thinking=animation / waiting=blinking hourglass / idle/inactive=dot. Color changes green→yellow→red with progress |
| Usage (%) |
Progress toward compact threshold or overall context usage (configurable) |
| Progress bar |
Color-coded by usage level |
| Element |
Example |
| Model |
opus |
| Messages |
57 msgs |
| Compact remaining |
compact 429.9k |
| Agents |
9 agents |
| Cost estimate |
$18.27 |
| Sparkline |
Context usage over time (off by default) |
Hover Actions
| Button |
Action |
| Pin / Unpin |
Pin session to top of list |
| Copy ID |
Copy session ID to clipboard |
| JSONL |
Open transcript file |
Click a card to focus the Claude Code session.
| Element |
Example |
| Context usage |
Context: 455.8k / 1.0M (46%) |
| Compact progress |
Compact at 49%: 429.9k left (52%) |
| Messages |
Messages: 121 |
| Token breakdown |
Input: 12.3M · Output: 890.5k · Cache hit: 78% |
| Cost |
Cost: $18.2742 |
| Compact count |
Compacts: 5 (auto: 4 / manual: 1) |
| Agents |
Agents: 9 spawned (2 active) |
Overview Panel
Aggregated stats across all sessions.
| Item |
Description |
| Active |
Total sessions (with thinking/waiting breakdown) |
| Max |
Highest context usage across all sessions |
| Rate |
API rate limit usage (5h window) and reset time |
| Cost |
Total cost across all sessions |
| Compacts |
Total compact events |
| Cleanup |
Days until Claude Code auto-deletes the oldest session (cleanupPeriodDays, default 30) |
| Updated |
Most recent session update |
Agent Graph Panel
Shows what the running session has delegated: the Claude session on one row, with its subagents and Codex CLI runs listed underneath, indented under a left rule.
| Element |
Description |
| Session row |
Blue pulsing dot = thinking, yellow = waiting, gray = idle. Shows the model, how many children are running out of the total, and context usage. Click to focus the session |
| Subagent row |
Blue pulsing dot while active. Shows the agent's name, its own model, and time since last activity; the task description appears in the tooltip |
| Codex row |
Green badge while running, with the subcommand, model, and elapsed time. Click to open the rollout transcript |
Running children are listed first, then completed ones by most recent activity. Completed children stay listed rather than disappearing, capped at 8 rows in the sidebar and 30 in the editor view, with a +N more row for the remainder.
Codex CLI sessions are detected from rollout files under ~/.codex/sessions (or CODEX_HOME) and matched to Claude sessions by working directory. The sidebar shows the focused or most recently active session; the title bar icon opens a full-screen editor view that lays all active sessions out in columns. Selecting a session in the Sessions panel focuses it here.
This extension is not affiliated with, endorsed by, or sponsored by OpenAI. It only reads local log files generated by your own Codex CLI usage; no data is sent to any external service.
Sort & Filter
Available from the Sessions view title bar menu (...).
Sort Order
Sort is applied in 3 stages: pinned → status (thinking→waiting→idle→inactive) → selected sort key.
| Sort |
Description |
| Sort by Time |
Most recently updated first (default) |
| Sort by Usage |
Highest context usage first |
| Sort by Compact Proximity |
Fewest tokens until compact first |
Filter
| Filter |
Description |
| Show All |
All sessions (default) |
| Warning Only |
Sessions above warning threshold |
| Critical Only |
Sessions above critical threshold |
Model Filter (independent axis, AND-combined with the above)
| Filter |
Description |
| Model All |
Sessions of any model (default) |
| Model Fable |
Fable sessions only |
| Model Opus |
Opus sessions only |
| Model Sonnet |
Sonnet sessions only |
| Model Haiku |
Haiku sessions only |
Session Status
| Status |
Condition |
Display |
| thinking |
Active response generation or tool execution |
Animated icon |
| waiting |
Waiting for permission approval (hook detection only) |
Blinking hourglass |
| idle |
Recently active but not currently processing |
Static dot |
| inactive |
No updates for 15+ seconds |
Static dot |
Cost Estimate
Calculates estimated cost based on token usage in session transcripts.
Pricing (per 1M tokens, as of 2026-07-25)
| Model |
Input |
Cache Write 5m |
Cache Write 1h |
Cache Read |
Output |
| Fable |
$10.00 |
$12.50 |
$20.00 |
$1.00 |
$50.00 |
| Opus (5 / 4.6+) |
$5.00 |
$6.25 |
$10.00 |
$0.50 |
$25.00 |
| Opus (4.1/4) |
$15.00 |
$18.75 |
$30.00 |
$1.50 |
$75.00 |
| Sonnet |
$3.00 |
$3.75 |
$6.00 |
$0.30 |
$15.00 |
| Haiku |
$1.00 |
$1.25 |
$2.00 |
$0.10 |
$5.00 |
Limitations
- Estimates based on public pricing, not actual billing
- Does not reflect Max/Pro plan discounts
Hook Detection
Uses Claude Code's hook system for precise session lifecycle detection.
- Enable via
enableHookDetection setting (off by default)
- Automatically configures Claude Code hooks when enabled and removes them when disabled
- Without hooks, thinking/idle detection still works via transcript monitoring (waiting requires hooks)
Settings
Search claude code vitals in VS Code settings (Ctrl+,).
| Setting |
Default |
Description |
enableHookDetection |
false |
Enable hook-based status detection. Automatically configures hooks when enabled, removes when disabled. Requires Node.js |
warningThreshold |
75 |
Warning threshold (%) |
criticalThreshold |
95 |
Critical threshold (%) |
notificationLevel |
none |
Notification trigger level (none/warning/critical) |
pollInterval |
60 |
Reconcile polling interval in seconds (1-60). Regular updates are watcher-driven |
inactiveHours |
24 |
Hide sessions inactive for this many hours |
defaultSort |
time |
Default sort order (time/usage/compact) |
defaultFilter |
all |
Default filter (all/warning/critical) |
defaultModelFilter |
all |
Default model filter (all/fable/opus/sonnet/haiku) |
defaultGroup |
none |
Default grouping mode (none/project/status/custom) |
useEnvOauthToken |
false |
Use CLAUDE_CODE_OAUTH_TOKEN from the environment before disk credentials when fetching usage |
progressMode |
compact |
Progress bar basis: compact or context |
cardDisplay |
model, messages, compact, agents, cost ON |
Card metadata elements |
tooltipDisplay |
all ON |
Tooltip elements |
debugGraphStateFile |
(empty) |
Debug only: dump the Agent Graph input state to this JSON file on each refresh. User/machine scope only — it cannot be set per workspace, and the target file is overwritten |
License
MIT
