Agent Registry

The Agent Registry is where you register, manage, and monitor every AI agent operating under TheWARDN governance.

Overview

Every AI agent that submits actions to the governance pipeline must first be registered. The Agent Registry provides a central view of all registered agents, their current status, activity statistics, and configuration.

Registering a New Agent

To register an agent, click Register Agent and provide:

• Name -- A unique, human-readable identifier for the agent
• Description -- A brief description of what the agent does and its purpose

Once registered, the agent receives an agent ID that must be included in all governance requests.

TIP

Use descriptive names that make it easy to identify agents in audit logs and the live monitor. For example, customer-support-responder is more useful than agent-1.

Agent List

The registry displays all registered agents in a table with real-time statistics:

Column	Description
Name	Agent display name
Status	Current lifecycle state (Active, Paused, Blocked, Deregistered)
Governed	Total number of actions submitted for governance
Cleared	Number of actions that received a CLEARED verdict
Held	Number of actions sent to escrow
Blocked	Number of actions denied

Agent Detail View

Click any agent to open its detail view, which includes:

• Full activity history
• Verdict distribution breakdown
• Policy assignments
• Configuration overrides
• Recent governance events

Agent Controls

Pause / Resume

Pausing an agent prevents it from submitting new actions. Any actions submitted by a paused agent are automatically blocked. Resume the agent to restore normal operation.

Block

Blocking an agent is a stronger action than pausing. A blocked agent's actions are denied and a AGENT_BLOCKED violation is recorded. Use this when an agent is behaving in ways that require investigation.

Deregister

Deregistering an agent permanently removes it from active governance. Historical audit records for the agent are preserved.

Agent Lifecycle

Agents move through the following states:

Active --> Paused --> Active       (reversible)
Active --> Blocked --> Active      (reversible, after review)
Active --> Deregistered            (permanent)
Paused --> Blocked --> Active      (reversible)
Paused --> Deregistered            (permanent)
Blocked --> Deregistered           (permanent)

WARNING

Deregistration is permanent. The agent's audit history is retained, but it cannot be re-activated. You would need to register a new agent.

Agent-Scoped Policies

You can assign CHAM policies that apply only to a specific agent. These override or supplement the global policy set. Common use cases:

• Setting a higher confidence floor for a new or untrusted agent
• Restricting an agent to specific action types
• Applying rate limits to a high-volume agent

Per-Agent Configuration

Confidence Floor

Set a minimum confidence threshold for the agent. Actions submitted with confidence below this floor are automatically held or blocked, regardless of other policy evaluations.

Tier Override

Override the default tier mapping for a specific agent. For example, you might allow a trusted internal agent to execute send_email at Tier A (autonomous), while an external agent is restricted to Tier B (supervised) for the same action type.

Activity History

The activity history tab shows a chronological log of all actions the agent has submitted, including verdicts, timestamps, and any violations triggered. Use this to audit agent behavior over time.