Agent Detail Page

The Agent Detail page provides a comprehensive view of a single agent. You reach this page by clicking an agent card on the Overview page or by navigating directly to /dashboard/agents/:id.

Header

The agent detail header displays the agent's identity and status at a glance:

┌──────────────────────────────────────────────────────┐
│ ⬡  Project Agent              ACTIVE    slack        │
│    Persistent agent for codespar/codespar            │
└──────────────────────────────────────────────────────┘

Stat Cards

Five stat cards are displayed below the header, providing key metrics for this agent:

Uptime

• Visual: Donut chart showing uptime percentage
• Calculation: Derived from uptimeMs — the number of milliseconds the agent has been running since last restart
• Display: Percentage with hours/days in secondary text (e.g., "99.7% — 14d 6h")
• Color: Green above 99%, amber between 95-99%, red below 95%

Tasks

• Primary value: Total tasks executed by this agent
• Secondary: Tasks executed today (e.g., "8 today")
• Trend: Small delta indicator compared to previous day

Memory

• Primary value: Current heap memory usage in MB
• Visual: Simple bar indicating usage relative to the system limit
• Color: Green under 256MB, amber between 256-512MB, red above 512MB

Build Rate

• Primary value: Percentage of successful builds associated with this agent's projects
• Secondary: Sparkline showing build success trend over the last 7 days
• Fallback: When no build data is available (e.g., no CI configured), the card displays "N/A" with a muted explanation: "No build data available"

Last Active

• Primary value: Relative time since the agent last processed a message or event
• Display: Human-readable format (e.g., "3 minutes ago", "1 hour ago", "2 days ago")
• Color: Green if within the last 5 minutes, amber if within 1 hour, gray if older

24-Hour Activity Chart

Below the stat cards, a bar chart shows the agent's activity over the past 24 hours, broken down by hour. Each bar represents the number of events processed in that hour.

Events
  12 │          ██
  10 │       ██ ██ ██
   8 │    ██ ██ ██ ██ ██
   6 │ ██ ██ ██ ██ ██ ██ ██
   4 │ ██ ██ ██ ██ ██ ██ ██ ██
   2 │ ██ ██ ██ ██ ██ ██ ██ ██ ██
   0 └──────────────────────────────
     00  03  06  09  12  15  18  21  Now

The chart uses Signal Blue (#3B82F6) for the bars and updates on each polling cycle.

Autonomy Level Control

The autonomy section displays the agent's current autonomy level as an interactive horizontal bar with six segments (L0 through L5):

Autonomy Level
┌─────┬─────┬─────┬─────┬─────┬─────┐
│ L0  │ L1  │ L2  │ L3  │ L4  │ L5  │
│Pass.│Notif│Sugg.│Auto │Auto │Full │
│     │     │     │ Low │ Med │Auto │
└─────┴─────┴─────┴─────┴─────┴─────┘
                    ▲
              Current: L3 (Auto-Low)

Changing Autonomy

Clicking a different level segment opens a confirmation dialog:

Change Autonomy Level

Are you sure you want to change the autonomy level
from L3 (Auto-Low) to L4 (Auto-Med)?

At L4, the agent will auto-execute medium-risk actions
including bug fixes and PR reviews.

[Cancel]  [Confirm Change]

On confirmation, the dashboard sends a request to the backend:

POST /api/agents/:id/action
Content-Type: application/json
 
{
  "action": "set_autonomy",
  "params": {
    "level": 4
  }
}

The change is logged in the audit trail and reflected immediately in the autonomy bar.

Tabs

The agent detail page has three tabs below the autonomy section: Activity, Context & Memory, and Configuration.

Activity Tab

The Activity tab shows a chronological timeline of all events related to this agent. Each entry includes:

Example timeline:

● 2m ago   Sarah Chen       task.complete    "Added /api/health endpoint"       SUCCESS
● 5m ago   project-agent    task.start       "Executing: add health endpoint"   SUCCESS
● 8m ago   Sarah Chen       command.receive  "@codespar instruct add health…"   SUCCESS
● 15m ago  System           webhook.receive  "workflow_run: CI passed"          SUCCESS
● 22m ago  review-agent     review.complete  "PR #42: Low risk, 3 suggestions"  SUCCESS

The timeline supports infinite scroll — older events load automatically as you scroll down.

Hash Chain

Every audit entry includes a cryptographic hash that chains to the previous entry, forming an immutable audit log. The hash is visible on hover and can be verified against the backend for tamper detection.

Context & Memory Tab

The Context & Memory tab provides visibility into the agent's internal knowledge:

Vector Store Stats

Module Graph

A summary of the agent's understanding of the codebase structure:

• Files indexed — number of source files the agent has read
• Dependencies mapped — import/require relationships tracked
• Key modules — top 5 most-referenced modules

Team Knowledge

Information the agent has learned about the team:

• Team members — names and roles the agent has observed
• Coding conventions — patterns the agent has detected
• Review preferences — merge rules, required approvals, etc.

CI/CD Patterns

Build and deployment intelligence:

• Average build time — based on observed workflow runs
• Common failure points — frequently failing tests or steps
• Deploy frequency — how often deploys happen

Configuration Tab

The Configuration tab displays and allows editing of agent-specific settings:

Execution Constraints

Alert Configuration

Toggle switches for notification preferences:

Each toggle immediately updates the agent configuration via the API.

Agent Actions

At the top right of the page, two action buttons are available:

Suspend / Resume

• Suspend: Pauses the agent. It stops processing messages and events but retains its state. The status badge changes to Paused (amber).
• Resume: Restarts a paused agent. It begins processing messages again from where it left off.

POST /api/agents/:id/action
{
  "action": "suspend"  // or "resume"
}

Restart

Forces a full restart of the agent — shuts it down and reinitializes it with its current configuration. Useful when an agent is in an error state or behaving unexpectedly.

POST /api/agents/:id/action
{
  "action": "restart"
}

A confirmation dialog is shown before restarting:

Restart Agent

Are you sure you want to restart "Project Agent"?
The agent will be briefly unavailable during restart.

[Cancel]  [Restart]

Data Flow

The agent detail page polls two endpoints:

Next Steps

• Audit Trail — view the full event log across all agents
• Settings — configure channels, team, and policies
• Autonomy Levels — learn more about the L0-L5 system