Skip to content

Agent DNA

1. Agent DNA

Agent DNA is the agent registry and lifecycle management hub. Every AI agent operating under Theseus Guard must have a registered DNA profile that defines its identity, capabilities, risk posture, and governance assignment.

1.1 Agent List & Catalog

The Agent DNA page opens to a searchable, filterable catalog of all registered agent profiles in the tenant. Each row displays:

  • Agent Name — The human-readable name assigned to the agent
  • Profile Key — A machine-readable unique identifier (e.g., finance-assistant-v2)
  • Status — Whether the agent is active, inactive, or pending
  • Guard Assignment — Which Guard is currently protecting this agent
  • Risk Profile — Classified as low, medium, or high based on L1 score thresholds
  • LLM Provider & Model — The AI provider and model the agent is configured to use
  • Department — The organizational unit this agent belongs to

Filters available: - Status (Active / Inactive / All) - Guard assignment - LLM provider - Risk profile

1.2 Agent Detail (Side Sheet)

Clicking any agent in the catalog opens a side sheet with the complete agent profile. This includes:

Identity Tab

  • Full agent name, profile key, tenant ID
  • Assigned department and subdepartment
  • Agent type (e.g., conversational, tool-calling, autonomous)
  • Creation and last-updated timestamps

Guard Assignment Tab

  • Currently active Guard name and domain
  • Guard configuration override options (available to admin/engineer roles)

Keys Tab

  • List of API keys associated with this agent
  • Key status, masked preview, last used date
  • Links to rotate or revoke individual keys

Compliance Tab

  • Current compliance status against tenant policy
  • Grace period countdown (if applicable)
  • List of active policy violations

1.3 Deploy New Agent (Wizard)

Users with admin or engineer roles can launch a multi-step wizard to register a new agent:

Step 1 — Identity - Set the agent name, description, and profile key - Select the agent type and assign a department

Step 2 — Model & Provider - Select the LLM provider (e.g., OpenAI, Anthropic, Azure) - Select the specific model - Configure fallback providers (if tenant policy allows)

Step 3 — Guard Assignment - Select an existing Guard from the catalog to govern this agent - Optionally configure guard domain overrides

Step 4 — Capabilities & Tools - Toggle allowed capabilities (RAG, web search, MCP, tool-calling) - Define allowed and blocked tools or MCP servers

Step 5 — Review & Deploy - Summary of all configuration choices - Submit to create the agent and generate an initial API key

Note: The auditor role cannot create or modify agents. The viewer role has read-only access to the catalog.

1.4 Edit & Retire Agents

Existing agents can be edited to update their guard assignment, capabilities, or department. Agents can also be set to inactive or permanently deleted, subject to the tenant's Agent Deletion Policy (configured in Admin → Agent Management).