Menu
Real-time agent monitoring with health scoring, cost tracking, and web dashboard
Install with Codex or Claude Copy this prompt, paste it into Codex, Claude, or another assistant, and let it review the skill page and install it for you.
Based on SOC occupation classification
Manage projects on Nulab Backlog โ issues, wikis, milestones, Git, sprints for Japan's top PM tool
Automate Japan's
Unified Gmail, Calendar, Drive, Sheets, and Docs management from Claude Code
Task and project management on Jooto โ create tasks, manage boards, automate workflows via PR TIMES' tool
Attendance and time tracking automation โ daily/monthly data, payroll sync via Japan's leading attendance system
Build, query, and automate Cybozu kintone apps โ Japan's leading no-code business platform
| name | Agent Dashboard |
| description | Real-time agent monitoring with health scoring, cost tracking, and web dashboard |
| author | hanabi-jpn |
| version | 1.0.0 |
| tags | ["monitoring","dashboard","analytics","cost-tracking","observability"] |
โญโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโฎ
โ โ
โ ๐ A G E N T D A S H B O A R D ๐ โ
โ โ
โ โโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโ โ
โ โ Health: โโโโโโโโโโ 82/100 โ โ
โ โ Tasks: 47 โ 3 โ 2 โ โ โ
โ โ Cost: $4.23 today โ โ
โ โ โโโโโโโโโโโโโโโโโโโโโโโโโ โ โ
โ โ ๐ โฑโฒ โฑโฒ โฑโฒ โ โ
โ โ โฑ โฒโฑ โฒโฑ โฒ trend โ โ
โ โโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโ โ
โ โ
โ "See everything. Miss nothing." โ
โฐโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโฏ
๐ Real-Time ๐ฐ Cost Tracking ๐ Alerts ๐ Web Export v1.0.0
Real-time monitoring and analytics for OpenClaw agents. Track performance, costs, task completion, and skill health with terminal and web dashboards.
Author: hanabi-jpn | Version: 1.0.0 | License: MIT
Tags: monitoring analytics dashboard performance cost-tracking
Agent Dashboard gives you full visibility into your OpenClaw agent's operations. Track health scores, task metrics, API costs, skill performance, and session history โ all in beautiful terminal displays or exportable web dashboards.
โโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโ
โ AGENT DASHBOARD ARCHITECTURE โ
โ โ
โ โโโโโโโโโโโ โโโโโโโโโโโโ โโโโโโโโโโโโโ โ
โ โ COLLECT โโโถโ ANALYZE โโโถโ VISUALIZE โ โ
โ โ Metrics โ โ Trends โ โ Terminal โ โ
โ โโโโโโโโโโโ โโโโโโโโโโโโ โโโโโโโโโโโโโ โ
โ โ โ โ
โ โโโโโโโโโโโ โโโโโโโโโโโโ โโโโโโโโโโโโโ โ
โ โ STORE โ โ ALERT โ โ EXPORT โ โ
โ โ JSONL โ โ Threshld โ โ HTML/JSON โ โ
โ โโโโโโโโโโโ โโโโโโโโโโโโ โโโโโโโโโโโโโ โ
โโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโ
You are an agent equipped with Agent Dashboard. Track and report on agent performance when requested.
After every significant action, log metrics to .agent-dashboard/metrics/:
Health Metrics (health.jsonl):
{
"timestamp": "2026-03-01T14:30:00Z",
"session_id": "sess-abc123",
"response_time_ms": 2300,
"tokens_input": 4500,
"tokens_output": 1200,
"tool_calls": 3,
"errors": 0,
"task_completed": true
}
Cost Tracking (costs.jsonl):
{
"timestamp": "2026-03-01T14:30:00Z",
"provider": "anthropic",
"model": "claude-sonnet-4-20250514",
"tokens_input": 4500,
"tokens_output": 1200,
"cost_usd": 0.0234
}
Skill Usage (skills.jsonl):
{
"timestamp": "2026-03-01T14:30:00Z",
"skill": "summarize-pro",
"command": "summarize",
"response_time_ms": 2300,
"success": true,
"tokens_used": 5700
}
health = (
error_rate_score ร 0.30 +
task_completion_score ร 0.30 +
response_quality_score ร 0.20 +
efficiency_score ร 0.20
)
where:
error_rate_score = max(0, 100 - (errors/total ร 200))
task_completion_score = (completed/attempted) ร 100
response_quality_score = 100 - (corrections/responses ร 200)
efficiency_score = max(0, 100 - (avg_tokens/1000 ร 10))
dashboard โ Terminal dashboard:
โโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโ
โ Agent Dashboard v1.0 โ
โ โโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโฃ
โ Health: โโโโโโโโโโ 82/100 Uptime: 99.2% โ
โ Tasks: 47 completed โ 3 failed โ 2 active โ
โ Cost: $4.23 today โ $28.91 this week โ
โ Tokens: 1.2M input โ 340K output โ
โ โโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโฃ
โ Skill Performance โ Calls โ Errs โ Avg ms โ
โ summarize-pro โ 23 โ 0 โ 2,300 โ
โ fx-trader-pro โ 18 โ 1 โ 4,700 โ
โ brain-trust โ 12 โ 0 โ 8,100 โ
โ humanize-ai-pro โ 8 โ 0 โ 1,900 โ
โ skill-guardian โ 5 โ 0 โ 12,400 โ
โ โโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโฃ
โ Cost Breakdown (Today) โ
โ Anthropic: $2.89 โโโโโโโโโโโโโโโโโ 68% โ
โ Google: $0.84 โโโโโโโโโโโโโโโโโ 20% โ
โ OpenAI: $0.50 โโโโโโโโโโโโโโโโโ 12% โ
โ โโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโฃ
โ Recent Activity โ
โ 14:32 โ
Analyzed EUR/USD M15 โ
โ 14:28 โ
Generated summary for report.pdf โ
โ 14:25 โ API timeout on Gemini (retried โ ok) โ
โ 14:20 โ
Humanized blog post (score 73%โ12%) โ
โ 14:15 โ
Brain Trust standup completed โ
โโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโ
dashboard health โ Detailed health breakdown
Output:
โโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโ
โ Agent Health โ Detailed Breakdown โ
โ โโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโฃ
โ Overall Health Score: 82/100 โ
โ โโโโโโโโโโโโโโโโโโโโ 82% โ
โ โ
โ Component Scores: โ
โ โโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโ โ
โ Error Rate: 92/100 (2.1% error rate) โ
โ Weight: 30% โ Contribution: 27.6 โ
โ โ
โ Task Completion: 88/100 (47/52 completed) โ
โ Weight: 30% โ Contribution: 26.4 โ
โ โ
โ Response Quality: 75/100 (5 corrections in 52) โ
โ Weight: 20% โ Contribution: 15.0 โ
โ โ
โ Efficiency: 65/100 (avg 3,500 tokens/task) โ
โ Weight: 20% โ Contribution: 13.0 โ
โ โ
โ Trend: โฒ +4 points from yesterday (78 โ 82) โ
โ Recommendation: Response quality can improve โ โ
โ review common correction patterns. โ
โโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโ
dashboard tasks [--period day|week|month] โ Task analytics:
Output:
โโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโ
โ Task Analytics โ This Week โ
โ โโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโฃ
โ Completed: 47 โ Failed: 3 โ Active: 2 โ
โ Success Rate: 94.0% (โฒ +2.1% from last week) โ
โ Avg Completion Time: 45 seconds โ
โ โ
โ Daily Breakdown: โ
โ Mon: โโโโโโโโโโโโ 12 tasks (100% success) โ
โ Tue: โโโโโโโโโโโ 10 tasks (90% success) โ
โ Wed: โโโโโโโโโโ 8 tasks (87.5% success) โ
โ Thu: โโโโโโโโโโโ 11 tasks (100% success) โ
โ Fri: โโโโโโโโโโ 9 tasks (88.9% success) โ
โ Sat: โโโโโโโโโโ 2 tasks (100% success) โ
โ โ
โ Most Common Task Types: โ
โ 1. Code analysis/editing: 18 (34%) โ
โ 2. FX trading operations: 12 (23%) โ
โ 3. Content generation: 8 (15%) โ
โ 4. File search/navigation: 7 (13%) โ
โ 5. Other: 7 (13%) โ
โโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโ
dashboard skills โ Skill performance ranking:
Output:
โโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโ
โ Skill Performance Ranking โ
โ โโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโฃ
โ Skill โ Calls โ Errs โ Avg ms โ Tokens โ
โ โโโโโโโโโโโโโโโโโโโโผโโโโโโโโผโโโโโโโผโโโโโโโโโผโโโโโโโโโโ โ
โ summarize-pro โ 23 โ 0 โ 2,300 โ 131,100 โ
โ fx-trader-pro โ 18 โ 1 โ 4,700 โ 84,600 โ
โ brain-trust โ 12 โ 0 โ 8,100 โ 97,200 โ
โ humanize-ai-pro โ 8 โ 0 โ 1,900 โ 45,600 โ
โ skill-guardian โ 5 โ 0 โ 12,400 โ 62,000 โ
โ context-slim โ 3 โ 0 โ 800 โ 2,400 โ
โ agent-dashboard โ 2 โ 0 โ 200 โ 1,200 โ
โ โ
โ Fastest: context-slim (800ms avg) โ
โ Slowest: skill-guardian (12,400ms avg) โ
โ Most tokens: summarize-pro (131,100 total) โ
โ โ
โ Unused Skills (installed but never called): โ
โ โข byterover (installed 14 days ago) โ
โ โข capability-evolver (installed 21 days ago) โ
โโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโ
dashboard cost [--period day|week|month] โ Cost analytics:
Output:
โโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโ
โ Cost Analytics โ This Month โ
โ โโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโฃ
โ Total: $28.91 (1 day into March) โ
โ Projected: $867.30 /month (at current rate) โ
โ โ
โ By Provider: โ
โ Anthropic: $19.84 โโโโโโโโโโโโโโโโ 69% โ
โ Google: $5.78 โโโโโโโโโโโโโโโโ 20% โ
โ OpenAI: $3.29 โโโโโโโโโโโโโโโโ 11% โ
โ โ
โ By Skill: โ
โ brain-trust: $9.72 (34%) โ
โ summarize-pro: $6.55 (23%) โ
โ fx-trader-pro: $5.20 (18%) โ
โ skill-guardian: $3.89 (13%) โ
โ other: $3.55 (12%) โ
โ โ
โ Cost per Task: $0.56 avg โ
โ Cheapest task: $0.02 (file search) โ
โ Most expensive: $4.82 (full security audit) โ
โ โ
โ Daily Trend: โ
โ Feb 24: $3.21 โ Feb 25: $4.89 โ Feb 26: $3.78 โ
โ Feb 27: $5.12 โ Feb 28: $4.23 โ Mar 01: $4.23 โ
โโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโ
dashboard alerts โ Show active alerts
Output:
โโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโ
โ Active Alerts โ
โ โโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโฃ
โ โ
โ โ ๏ธ [Mar 01 14:25] Daily cost ($5.23) exceeded โ
โ threshold ($5.00). Current: $5.23 โ
โ Triggered by: brain-trust meeting ($1.82) โ
โ โ
โ โ ๏ธ [Mar 01 14:25] fx-trader-pro error count: 3 โ
โ threshold: 3. Last error: API timeout on โ
โ candle fetch for GBP_NZD โ
โ โ
โ Resolved Today: โ
โ โ
[Mar 01 10:15] Health score dropped to 58 โ
โ Recovered to 82 after 2 successful tasks โ
โ โ
โ Alert Configuration: โ
โ error-rate: 10% โ daily-cost: $5 โ health: 60 โ
โ monthly-cost: $50 โ fx-trader-pro errors: 3 โ
โโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโ
dashboard alert set <type> <threshold> โ Configure alerts:
error-rate 10 -- Alert if error rate exceeds 10%daily-cost 5 -- Alert if daily cost exceeds $5monthly-cost 50 -- Alert if monthly cost exceeds $50skill-error <skill> 3 -- Alert if skill fails 3+ timeshealth 60 -- Alert if health score drops below 60Output:
โ
Alert configured: daily-cost โ $8.00
Current Alert Thresholds:
โโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโ
error-rate: 10% (current: 2.1%) โ
daily-cost: $8.00 (current: $5.23) โ
monthly-cost: $50.00 (current: $28.91) โ
health: 60 (current: 82) โ
fx-trader-pro: 3 errors (current: 1) โ
Saved to: .agent-dashboard/alerts/config.json
dashboard report [daily|weekly|monthly] โ Generate markdown report:
Output:
๐ Weekly Report generated: .agent-dashboard/reports/2026-03-01.md
# Agent Dashboard Weekly Report โ 2026-03-01
## Summary
- Tasks completed: 47 (94% success rate)
- Total cost: $28.91 ($4.13/day avg)
- Health score: 82 (โฒ +4 from last week)
- Most active skill: summarize-pro (23 calls)
- API calls: 1,247 (Anthropic: 842, Google: 312, OpenAI: 93)
## Highlights
- Zero errors on summarize-pro (23 consecutive successes)
- Health score trending upward 3 days straight
- Cost per task decreased 12% from last week
## Issues
- fx-trader-pro: 1 API timeout (auto-recovered)
- Response quality dip on Wednesday (3 corrections)
## Cost Forecast
- Projected March total: $124.00 (within $150 budget)
Report saved to .agent-dashboard/reports/2026-03-01.md (5.1 KB)
dashboard --web โ Generate HTML dashboard:
Output:
โโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโ
โ Web Dashboard Generated โ
โ โโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโฃ
โ โ
โ Output: .agent-dashboard/web/dashboard.html โ
โ Size: 142 KB (self-contained, no dependencies) โ
โ โ
โ Includes: โ
โ โ
Health score gauge with trend chart โ
โ โ
Task completion line graph (7 days) โ
โ โ
Cost breakdown pie chart by provider โ
โ โ
Skill performance table (sortable) โ
โ โ
Recent activity timeline โ
โ โ
Alert history panel โ
โ โ
Dark mode toggle โ
โ โ
Mobile responsive layout โ
โ โ
โ To view: open .agent-dashboard/web/dashboard.html โ
โ To serve: python3 -m http.server 8080 โ
โ (auto-refresh every 60s) โ
โโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโ
dashboard replay [--session <id>] โ Session replay:
Output:
โโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโ
โ Session Replay: sess-abc123 โ
โ Started: 2026-03-01 14:00 UTC โ
โ Duration: 47 minutes โ
โ โโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโฃ
โ โ
โ 14:00:12 [START] Session initialized โ
โ 14:01:30 [fx-trader-pro] fx status โ
โ โ
2,300ms โ 4,500 in / 1,200 out โ
โ 14:05:45 [fx-trader-pro] fx scan โ
โ โ
12,400ms โ 28,000 in / 8,400 out โ
โ 14:08:22 [fx-trader-pro] fx signal GBP_JPY โ
โ โ
4,700ms โ 8,200 in / 2,100 out โ
โ 14:12:00 [fx-trader-pro] fx trade GBP_JPY buy โ
โ โ
3,100ms โ 6,400 in / 1,800 out โ
โ 14:20:10 [brain-trust] bt meeting standup โ
โ โ
8,100ms โ 12,000 in / 6,200 out โ
โ 14:25:33 [summarize-pro] summarize report.pdf โ
โ โ TIMEOUT โ Retried โ โ
4,200ms โ
โ 14:32:15 [context-slim] slim optimize โ
โ โ
800ms โ 2,400 in / 600 out โ
โ 14:47:00 [END] Session complete โ
โ โ
โ Session totals: 7 actions โ 1 error โ $4.23 cost โ
โโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโ
dashboard export <json|csv|html> โ Export all data
Output:
โโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโ
โ Data Export โ JSON โ
โ โโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโฃ
โ โ
โ Exporting all metric data... โ
โ โ
โ Files generated: โ
โ โ
health-metrics.json (1,247 entries, 2.1 MB) โ
โ โ
task-records.json ( 312 entries, 840 KB) โ
โ โ
skill-usage.json ( 189 entries, 420 KB) โ
โ โ
cost-tracking.json ( 892 entries, 1.4 MB) โ
โ โ
session-data.json ( 42 sessions, 3.2 MB)โ
โ โ
alert-history.json ( 28 entries, 64 KB) โ
โ โ
โ Total: 6 files, 8.0 MB โ
โ Output: .agent-dashboard/exports/2026-03-01/ โ
โ โ
โ Format: JSON (UTF-8, pretty-printed) โ
โ Compatible with: jq, Elasticsearch, Grafana, etc. โ
โโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโ
dashboard reset โ Clear all tracking data (confirmation required)
Output:
โโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโ
โ Dashboard Reset โ
โ โโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโฃ
โ โ
โ โ ๏ธ This will permanently delete all tracking data. โ
โ โ
โ Data to be cleared: โ
โ Health metrics: 1,247 entries (2.1 MB) โ
โ Task records: 312 entries (840 KB) โ
โ Skill usage: 189 entries (420 KB) โ
โ Cost tracking: 892 entries (1.4 MB) โ
โ Session replays: 42 sessions (3.2 MB) โ
โ Alert history: 28 entries (64 KB) โ
โ โ
โ Tip: Run `dashboard export json` first to backup. โ
โ โ
โ Confirm reset? (yes/no): yes โ
โ โ
โ โ
All tracking data cleared. โ
โ Alert configuration preserved. โ
โ Health score reset to 100 (fresh start). โ
โโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโ
{
"anthropic": {
"claude-opus-4-20250514": {"input": 15.0, "output": 75.0},
"claude-sonnet-4-20250514": {"input": 3.0, "output": 15.0},
"claude-haiku-3-5-20241022": {"input": 0.80, "output": 4.0}
},
"openai": {
"gpt-4o": {"input": 2.5, "output": 10.0},
"gpt-4o-mini": {"input": 0.15, "output": 0.60}
},
"google": {
"gemini-2.5-pro": {"input": 1.25, "output": 10.0},
"gemini-2.5-flash": {"input": 0.15, "output": 0.60}
}
}
Prices per 1M tokens in USD. Updated periodically.
.agent-dashboard/
โโโ metrics/
โ โโโ health.jsonl # Health metrics per action
โ โโโ tasks.jsonl # Task records
โ โโโ skills.jsonl # Skill usage records
โ โโโ costs.jsonl # Cost tracking
โโโ alerts/
โ โโโ config.json # Alert thresholds
โโโ reports/
โ โโโ {date}.md # Generated reports
โโโ sessions/
โ โโโ {session-id}.jsonl # Session replay data
โโโ config.json # Dashboard settings
When enabled, alerts fire when thresholds are crossed:
.agent-dashboard/alerts/history.jsonlโ ๏ธ ALERT: Daily cost ($5.23) exceeded threshold ($5.00)
โ ๏ธ ALERT: Health score dropped to 58 (threshold: 60)
Agent Dashboard handles metric collection, export, and notification failures gracefully to ensure monitoring continuity.
| Scenario | Handling |
|---|---|
| Unable to log health metrics after an action | Write to a local buffer (/tmp/agent-dashboard-buffer.jsonl). Retry writing to .agent-dashboard/metrics/health.jsonl on next successful action. Never block the user's operation to log metrics. |
| Metrics file locked by another process | Wait up to 2 seconds for lock release. If still locked, buffer the entry and append on next write. Use file-level locking to prevent corruption. |
| Disk full โ cannot write metrics | Switch to in-memory-only mode. Display a warning: "Dashboard metrics storage full. Monitoring continues in memory only." Suggest dashboard reset or clearing old data. |
| Token/cost data unavailable from provider | Record the action with cost_usd: null and tokens: null. Mark the entry as "incomplete" for later reconciliation. Dashboard totals exclude null entries with a footnote. |
| Skill usage tracking fails for unknown skill | Log the event with the raw skill name. Create a new entry in skills.jsonl. Do not discard usage data for unrecognized skills. |
| Scenario | Handling |
|---|---|
| HTML export fails (template rendering error) | Fall back to plain JSON export. Display the raw data with a note that HTML generation failed. Log the template error for debugging. |
| CSV export with special characters in data | Escape all fields with double-quote wrapping. Handle Unicode, commas, and newlines in field values. Use UTF-8 BOM for Excel compatibility. |
| Export file too large (>50MB) | Split into multiple files by date range (monthly chunks). Alert user with file count and total size. |
| Web dashboard generation fails | Provide the static terminal dashboard as fallback. Log the error. Common cause: corrupted metrics data โ suggest dashboard reset with date range to clean specific periods. |
| Scenario | Handling |
|---|---|
| Alert threshold crossed but notification cannot be displayed | Buffer the alert in .agent-dashboard/alerts/pending.jsonl. Display all pending alerts at the start of the next agent interaction. |
| Alert config file corrupted | Fall back to default thresholds (error-rate: 10%, daily-cost: $10, health: 50). Warn user that custom alert settings need reconfiguration. |
| Alert storm (>10 alerts in 1 minute) | Consolidate into a single summary alert: "Multiple alerts triggered (12 events in last minute). Run dashboard alerts for details." Prevents alert fatigue. |
| False positive alert (metric spike then recovery) | Alerts fire on threshold crossing, not sustained state. Include "current value" and "threshold" in every alert so the user can assess. Historical alerts are always available in alerts/history.jsonl. |
| Scenario | Handling |
|---|---|
| Session data incomplete (agent crashed mid-session) | Display available data up to the last recorded event. Mark the session as "incomplete" in the replay timeline. |
| Session ID not found | List available session IDs with dates. Suggest closest match if the ID looks like a typo. |
dashboard export json periodically to create backups.dashboard reset --before <date> clears old data while preserving recent metrics.| Feature | Agent Dashboard | Manual Logging (print/console) | Console Output (Default) | Langfuse | Weights & Biases |
|---|---|---|---|---|---|
| Setup | Zero config โ installs with skill | None needed | None needed | Account + SDK integration | Account + SDK integration |
| Health Score | Composite 0-100 score (error rate, completion, quality, efficiency) | None | None | Trace-level scoring | Custom metrics only |
| Cost Tracking | Automatic per-provider, per-model, per-skill breakdown | Manual calculation | None | Automatic with SDK | Custom logging required |
| Token Tracking | Automatic input/output per action | Manual if coded | None | Automatic with SDK | Custom logging required |
| Skill-Level Metrics | Per-skill calls, errors, avg response time, token usage | Not practical manually | None | Per-trace breakdown | Custom tagging |
| Alerting | Built-in threshold alerts (cost, health, error rate) | None | None | Custom alerts (paid plan) | Custom alerts |
| Terminal Dashboard | Rich ASCII dashboard with real-time data | N/A | Raw text only | Web UI only | Web UI only |
| Web Dashboard | Self-contained HTML export (offline capable) | N/A | N/A | Cloud-hosted web UI | Cloud-hosted web UI |
| Session Replay | Full action timeline per session | Not feasible manually | Scroll through terminal | Trace replay (paid plan) | Not available |
| Data Storage | Local JSONL files (private, no cloud dependency) | Scattered log files | Ephemeral (lost on close) | Cloud-hosted (their servers) | Cloud-hosted (their servers) |
| Privacy | 100% local โ no data leaves your machine | Local | Local | Data sent to Langfuse cloud | Data sent to W&B cloud |
| Cost of Tool | Free (MIT license) | Free | Free | Free tier limited / $59+/mo | Free tier limited / $50+/mo |
| OpenClaw Integration | Native โ tracks all OpenClaw skills automatically | None | None | Requires custom integration | Requires custom integration |
Q: How much does Agent Dashboard cost? A: The skill itself is completely free (MIT license). It tracks your existing API costs but does not add any additional API calls or fees. All data is stored locally.
Q: What data does Agent Dashboard collect? A: It tracks response times, token counts (input/output), tool calls, error counts, task completion status, and API costs. It estimates costs based on built-in pricing tables for major providers (Anthropic, OpenAI, Google). No personal data or conversation content is stored โ only operational metrics.
Q: How long is metric data retained?
A: Indefinitely by default. Data is stored in JSONL files that grow over time. Use dashboard reset --before <date> to clear old data, or dashboard export to archive before clearing. For long-running agents, consider monthly exports and resets to keep file sizes manageable.
Q: Does it affect agent performance? A: Negligible impact. Metric logging is append-only to JSONL files (microseconds per write). The dashboard display itself uses only data already collected โ no additional API calls are made. Health score calculation is lightweight arithmetic.
Q: Can I set up alerts for specific skills?
A: Yes. Use dashboard alert set skill-error <skill-name> <threshold> to alert when a specific skill exceeds a failure count. You can also set global alerts for error rate, daily cost, monthly cost, and health score.
Q: How do I integrate with external monitoring tools?
A: Use dashboard export json to get all metrics as JSON. This can be piped to any external tool, ingested by Grafana, or processed by custom scripts. The JSONL format is compatible with standard log processing pipelines (jq, Elasticsearch, etc.).
Q: Is my data sent to any cloud service?
A: No. Agent Dashboard is 100% local. All data is stored in .agent-dashboard/ in your project directory. No telemetry, no cloud sync, no external API calls. The web dashboard export generates a self-contained HTML file that works offline.
Q: How accurate are the cost estimates?
A: Cost estimates use built-in pricing tables that are updated periodically. They are accurate to within 5% for standard API usage. Variations can occur if your provider offers custom pricing, volume discounts, or if prices have changed since the last pricing table update. You can update pricing in .agent-dashboard/config.json.
Q: Can I track multiple agents or projects?
A: Each project has its own .agent-dashboard/ directory with independent metrics. To get a cross-project view, use dashboard export json from each project and aggregate externally. A multi-project dashboard feature is planned for a future version.
Q: What happens if I forget to check the dashboard? A: Alerts are your safety net. Set thresholds for the metrics you care about most (daily cost, error rate, health score), and Agent Dashboard will notify you inline whenever a threshold is crossed. Pending alerts are displayed at the start of each new session.