Docs/Core Concepts/Agent Monitor

Agent Monitor

Gain complete visibility over every AI agent in your organization. Discover what's running, protect the agents you care about, and detect changes or anomalies the moment they happen.


Overview

As organizations adopt AI agents across multiple platforms — n8n workflows, Make scenarios, LangChain applications, custom scripts — it becomes impossible to track what's running without a dedicated system.

The Trappr Agent Monitor provides a unified view of all agents registered with your workspace. Each agent has a lifecycle: it's discovered, optionally protected, and continuously monitored for behavioral changes.

Agent states

StateMeaning
ACTIVEAgent is registered, protected, and behaving as expected
PENDINGAgent is registered but not yet verified — first contact not received
TAMPEREDAgent fingerprint has changed — possible modification or compromise
DEACTIVATEDAgent has been deactivated after a tamper event; access revoked

Discover mode

When you first register an agent with Trappr, it enters Discover mode. In this mode, Trappr observes the agent's traffic and behavior without enforcing any policies. This is the recommended starting point — it gives you a complete picture of what your agent does before you lock it down.

Discover mode is fully passive. No requests are blocked, no alerts are fired (unless a canary is triggered). Use it for 24–72 hours to establish a behavioral baseline.

Protect mode

Once you've reviewed the baseline, switch the agent to Protect mode. Trappr begins enforcing the policies you've configured — DLP masking, canary monitoring, and optionally request blocking via Enforcement Mode.

Register Agent
dashboard / API
Discover
observe baseline
Protect
enforce policies
Monitor
continuous alerting

Registering an agent

1
Create an agent in the dashboard

Go to Dashboard → Agents → Register agent. Provide a name and the platform it runs on (n8n, Make, LangChain, custom, etc.). Trappr generates a unique agent ID (cht_agent_XXXX) and an agent API key (cht_ak_XXXX).

2
Add the agent key to your agent

Store the agent key as an environment variable or secret in your agent's runtime. Route all LLM calls through the Gateway using this key. The key identifies which agent each request belongs to.

3
Verify the first contact

Trigger your agent once. In the dashboard, the agent should move from PENDING to ACTIVE. You'll see the first execution record appear under Agent → Executions.

4
Switch to Protect mode when ready

After observing the baseline in Discover mode, click Enable protection on the agent detail page. Trappr begins enforcing your security policies on all subsequent executions.

Tamper detection

Trappr continuously verifies that your registered agents haven't been modified. If an agent's behavior or identity deviates from its registered profile, it's flagged as TAMPEREDand an alert is fired immediately.

In Protect mode with enforcement enabled, a tampered agent is automatically deactivated — its API key is revoked and further requests are blocked until you investigate and re-register.

A TAMPERED state means Trappr detected a change that doesn't match the agent's registered profile. This could indicate a supply-chain compromise, an injected step, or unauthorized modification. Investigate before re-activating.

Execution history

Every agent execution is tracked end-to-end. Under Agent → Executions, you can see:

  • Start and end time of each execution
  • All LLM calls made during the execution
  • DLP events, canary triggers, and policy violations
  • Tool calls and external API calls
  • Token usage and estimated cost