Skip to main content

​
Viewing Query Events in the Dashboard

The Agent-CoreX Dashboard provides a comprehensive view of all tool retrieval queries, enabling you to monitor which tools your agents are selecting and understand retrieval patterns.

​
Getting Started

​
Prerequisites

  • ✅ Active Agent-CoreX account
  • ✅ Dashboard access at dashboard.agent-corex.com
  • ✅ At least one authenticated query (see below)

​
Note: Anonymous Queries

Query events require user authentication. If you’re making requests without an API key:
  • ✅ Tools will still be retrieved and returned
  • ❌ Query event will not be logged
  • ✓ Check backend logs instead

​
Accessing Query Events

​
Step 1: Navigate to Queries Page

  1. Log in to the dashboard
  2. Click Queries in the sidebar
  3. Select Query History
You’ll see a table of recent queries from your account.

​
Understanding the Query List

​
Columns

ColumnMeaning
QueryThe natural language query (e.g., “deploy to production”)
ToolsCount of tools retrieved
SourceOrigin: mcp (CLI), api (direct API), dashboard (UI)
TimeWhen the query was executed
StatusSuccess or failure indicator

​
Example

Query                        | Tools | Source | Time
─────────────────────────────┼───────┼────────┼──────────────
"deploy to production"       | 3     | mcp    | 2 min ago
"get github issues"          | 5     | api    | 5 min ago
"send slack notification"    | 2     | mcp    | 12 min ago

​
Viewing Query Details

​
Click on a Query

  1. Click any query row to open the Query Details Panel
  2. See full information:
    • Query text
    • Retrieved tools (all candidates)
    • Tool rankings (scores)
    • Execution time
    • Request source

​
Details Breakdown

​
Retrieved Tools Section

Shows all tools returned by semantic search, ranked by relevance: 
Tool Name               Score   Server      Status
────────────────────────────────────────────────────
deploy-aws             0.98    aws-mcp     ✓ Selected
send-slack-message     0.92    slack-mcp   ✓ Selected
create-github-issue    0.78    github-mcp  ○ Not used
notify-pagerduty       0.65    pd-mcp      ○ Not used
What do the icons mean?
  • ✓ Selected — Tool was actually used
  • ○ Not used — Tool was available but not selected
  • ⚠️ Failed — Tool was selected but execution failed

​
Scores Section

Detailed breakdown of ranking metrics: 
Tool: deploy-aws
├─ Relevance (semantic): 0.96
├─ Capability Match: 0.99
├─ Success Rate: 0.95
└─ Overall Score: 0.98
Interpreting scores:
  • 0.90-1.00 — Excellent match
  • 0.70-0.89 — Good match
  • 0.50-0.69 — Fair match
  • Below 0.50 — Poor match

​
Filtering & Searching

​
By Date Range

  1. Click Date Range filter
  2. Select preset (Last 24h, Last 7d, Last 30d) or custom dates
  3. Results update automatically

​
By Source

Click Source filter to show only:
  • All — Every query
  • MCP — CLI/MCP gateway queries
  • API — Direct API calls
  • Dashboard — Dashboard UI queries
Use case: Filter by “MCP” to see only queries from your agents.

​
By Tool

  1. Click Tools filter
  2. Select specific tools (e.g., “deploy-aws”, “send-slack”)
  3. Shows queries that retrieved those tools

​
By Query Text

Search for specific queries: 
Search: "deploy"
Results:
├─ "deploy to production"
├─ "deploy and notify"
└─ "deploy docker image"

​
Query Volume Over Time

View the Query Analytics tab: 
Chart: Queries per Day
      │
  20  │  ╱╲
      │ ╱  ╲     ╱╲
  10  │╱    ╲   ╱  ╲
      │      ╲ ╱    ╲
   0  └───────────────────
      Mon Tue Wed Thu Fri
What it shows:
  • Peak usage times
  • Daily query volume trends
  • Spikes in agent activity

​
Top Tools by Selection

See which tools are most frequently used: 
Tool Name          Selected Count  Percentage
─────────────────────────────────────────────
deploy-aws         87              23%
send-slack         64              17%
create-issue       52              14%
get-logs           48              13%
...
Use case: Identify which tools agents rely on most.

​
Top Queries

See most frequent natural language queries: 
Query                       Count
───────────────────────────────────
"deploy to production"      12
"notify team"               8
"create github issue"       7
"get error logs"            6
Use case: Understand what agents ask for most.

​
Debugging Tool Selection

​
Problem: “Wrong tools retrieved”

  1. Find the relevant query in history
  2. Open query details
  3. Check Retrieved Tools section
  4. Compare scores to understand ranking
Example: 
Query: "send a message to slack"

Tools Retrieved:
├─ send-slack-message (0.95) ✓ Selected ← Correct!
├─ deploy-aws (0.42)
└─ github-create-pr (0.38)

​
Problem: “Tool not retrieved at all”

  1. Open the query details
  2. Check if tool appears in Retrieved Tools
  3. If tool not present:
    • Tool may not exist in your connected servers
    • Server may be offline
    • Tool may not match the query semantically
  4. If tool is present but not selected:
    • Check the score (too low?)
    • Too many better tools were available
    • Tool filtering may have excluded it

​
Performance Insights

​
Slow Queries

If you see a query took >1 second:
  1. Click the query
  2. View Execution Time details
  3. Check Retrieved Tools count
  4. High count = more semantic search time (expected)

​
Successful vs Failed Execution

Some queries show:
  • Retrieved — Tools were found ✓
  • Executed — Tools were called ✓
  • Failed — Tool execution failed ⚠️
Check tool execution logs for failure reasons.

​
Export & Reporting

​
Export Query History

  1. Open Queries page
  2. Click ⬇️ Export
  3. Choose format:
    • CSV (for Excel/spreadsheets)
    • JSON (for analysis/APIs)
    • PDF (for reports)

​
Generate Report

  1. Click 📊 Report
  2. Select metrics:
    • Total queries
    • Tool selection trends
    • Source distribution
    • Query performance
  3. Choose date range
  4. Download PDF
Example report sections:
  • Executive Summary (query volume, top tools)
  • Trend Analysis (queries over time)
  • Tool Performance (selection frequency, success rates)
  • Recommendations (frequently used tools, failing queries)

​
Common Use Cases

​
1. Monitor Agent Activity

Goal: See what your deployed agent is doing Steps:
  1. Filter by Source: MCP (agent queries)
  2. View recent query history
  3. Check tool selection patterns

​
2. Debug Tool Selection Issues

Goal: Understand why wrong tool was selected Steps:
  1. Find the problematic query
  2. Open query details
  3. Review all retrieved tools and scores
  4. Check if better tool existed but wasn’t selected
  5. Report to team or adjust semantic search

​
3. Analyze Tool Popularity

Goal: Find which tools are most useful Steps:
  1. Go to Query Analytics
  2. View Top Tools by Selection
  3. Correlate with business goals
  4. Consider optimizing/expanding popular tools

​
4. Performance Tuning

Goal: Optimize for faster queries Steps:
  1. Monitor Execution Time trends
  2. Identify slow queries
  3. Check if high tool counts cause slowness
  4. Consider filtering or caching strategies

​
5. Cost Analysis

Goal: Track API usage costs Steps:
  1. Export query history (CSV)
  2. Sum queries by date
  3. Multiply by per-query cost
  4. Monitor spend over time

​
Keyboard Shortcuts

ShortcutAction
EnterOpen selected query details
EscClose details panel
↑↓Navigate query list
Ctrl+FSearch queries
Ctrl+EExport current view

​
Settings & Preferences

​
Auto-Refresh

Enable automatic refresh of query list:
  1. Click ⚙️ Settings
  2. Toggle Auto-refresh
  3. Choose interval (5s, 10s, 30s)
Useful when monitoring live agent activity.

​
Display Preferences

  1. Click ⚙️ Settings
  2. Show scores — Toggle detailed score display
  3. Show metadata — Toggle tool metadata
  4. Compact view — Reduce row height
  5. Dark mode — Toggle theme

​
Troubleshooting

​
”No query events appearing”

Checklist:
  • Using authenticated API key (not anonymous)
  • Queries are recent (check date filter)
  • Source filter isn’t too restrictive
  • Dashboard is connected to correct environment

​
”Scores look wrong”

Scores reflect:
  • Semantic similarity to your query
  • Capability match (does tool do what you need?)
  • Historical success rate (has it worked before?)
High score = tool should work well for your query.

​
”Tool isn’t appearing in retrieved tools”

Reasons:
  1. Tool may not exist in connected servers
  2. Server is offline/disconnected
  3. Tool name doesn’t semantically match your query
  4. Tool filtered by your search criteria

​
Next Steps

Query Events Concept

Deep dive into how query events work.

Tool Retrieval

Understand ranking and scoring.

API Reference

Programmatic access to query events.

Getting Started

Make your first API call.
Key Takeaway: The dashboard’s Query History provides complete visibility into which tools your agents are retrieving and selecting. Use it to monitor, debug, and optimize your agent systems. 🔍📊