> ## Documentation Index
> Fetch the complete documentation index at: https://vizmaxxing.com/llms.txt
> Use this file to discover all available pages before exploring further.

# Agent Therapy: AI Agent Behavioral Health Platform

> Agent Therapy monitors, diagnoses, and corrects behavioral issues in AI agents — keeping your systems reliable, predictable, and on-task.

Agent Therapy is a behavioral health platform for AI agents. If you're running LLM-powered agents in production — customer support bots, coding assistants, data pipelines, or autonomous workflows — you've likely encountered the frustrating reality: agents hallucinate, drift off-task, get stuck in loops, or silently fail in ways that are hard to detect and even harder to fix. Agent Therapy gives your team a systematic way to catch these issues early, understand what went wrong, and apply targeted corrections before they impact your users or your business.

## What Agent Therapy Does

Agent Therapy is built on three core pillars that work together across the full agent lifecycle:

**Monitor** — Agent Therapy continuously tracks your agents' behavioral health in real time. Every session generates a live Agent Health Score (0–100) that reflects how consistently, accurately, and safely your agent is operating. You see drift, anomalies, and degradation the moment they emerge — not after a user complains.

**Diagnose** — When something goes wrong, Agent Therapy automatically analyzes your agent's behavior logs to identify the root failure mode. Whether it's instruction drift, a hallucination burst, a repetitive loop, or a chain of failed tool calls, you get a clear, evidence-backed diagnostic report that tells you exactly what happened and why.

**Intervene** — Armed with a diagnosis, Agent Therapy's intervention library gives you targeted, non-destructive corrections to apply directly to your agent. From prompt corrections and constraint injections to behavioral resets and conversation pruning, interventions are designed to fix the specific issue without destabilizing the rest of your agent's behavior.

## Who Is It For?

Agent Therapy is built for anyone responsible for the reliability of AI agents in production:

* **AI Engineers** who build and maintain agent systems and need deep observability into runtime behavior — beyond what logs and traces alone can provide.
* **Product Teams** who own user-facing AI features and need confidence that agents are behaving as intended, without requiring deep ML expertise to investigate issues.
* **Platform & MLOps Teams** who operate fleets of agents at scale and need automated monitoring, alerting, and governance tools to maintain quality across many agents simultaneously.
* **Compliance & Trust & Safety Teams** who need auditable records of agent behavior to meet regulatory requirements or internal policy standards.

If your team is shipping agents to real users, Agent Therapy is for you.

## Key Capabilities

> **Behavioral Monitoring**
>
> Track real-time Agent Health Scores across all your agents. Surface behavioral anomalies, performance degradation, and safety violations the moment they occur — with full session-level detail.

> **Session Diagnostics**
>
> Automatically analyze behavior logs to detect known failure modes including instruction drift, hallucination bursts, repetitive loops, tool call failures, and context overflow.

> **Intervention Library**
>
> Apply targeted, evidence-based corrections to misbehaving agents. Choose from prompt corrections, constraint injections, behavioral resets, and conversation pruning — all non-destructive and reversible.

> **Alerts & Notifications**
>
> Configure threshold-based alerts that fire when an agent's health score drops, a specific failure mode is detected, or session error rates spike. Deliver alerts to Slack, PagerDuty, email, or any webhook.

> **Integration Ecosystem**
>
> Connect Agent Therapy to the frameworks and tools you already use — OpenAI, LangChain, LlamaIndex, Anthropic, and more — via native SDKs, a REST API, and pre-built middleware adapters.

> **Audit Trails**
>
> Maintain a tamper-evident log of every session, diagnostic finding, and intervention applied. Meet compliance requirements and give your team a full history of agent behavior over time.

## Next Steps

> **[Quickstart](https://vizmaxxing.com/quickstart.md)**
>
> Connect your first agent and run a diagnostic session in under 5 minutes.

> **[How It Works](https://vizmaxxing.com/how-it-works.md)**
>
> Dive deeper into the therapy lifecycle, health scoring model, and intervention system.


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.