# Agent Self-Reflection & Sentiment Scanner MCP for AI Agents MCP

> Agent Self-Reflection & Sentiment Scanner MCP for monitoring agent execution logs. It identifies self-correction markers and analyzes the emotional tone of an agent's reasoning to help you understand if it is actually making progress or just looping on errors.

## Overview
- **Category:** observability
- **Price:** Free
- **Endpoint:** https://edge.vinkius.com/vk_preview_F2yFzx4CchXScg2i4EwfRYqrArVJ0UEEfstzE52Z/mcp
- **Tags:** sentiment-analysis, self-reflection, agent-monitoring, log-analysis, error-detection

## Description

Checking agent logs usually means scrolling through endless terminal outputs or JSON files, looking for that one error string. It is a manual, exhausting way to find out why a task failed. This MCP changes that by turning raw execution logs into measurable data. You can see exactly how many times your agent realizes it made a mistake and restarts a step. It also lets you peek into the reasoning process to see if the agent is expressing frustration or confidence during a task. By connecting this to your workflow through Vinkius, you get a clear view of whether your agent is actually proceeding or stuck in a correction loop. No more manual scanning through thousands of lines of text to find where things went sideways.

## Tools

### analyze_sentiment
Checks the emotional tone of log entries to see if the agent is struggling or succeeding. It helps you identify shifts in confidence during a task.

### detect_reflection
Finds specific parts of a log where the agent expresses internal thoughts or reasoning. This makes it easy to audit the agent's decision-making process.

## Prompt Examples

**Prompt:** 
```
What happened in my last run?
```

**Response:** 
```
| Log Entry | Sentiment | Status |
| :--- | :--- | :--- |
| "I made a mistake..." | Negative | Correcting |
| "The task is complete." | Positive | Proceeding |

Your agent encountered one error and successfully recovered.
```

**Prompt:** 
```
Is my agent getting frustrated?
```

**Response:** 
```
Based on the recent logs, there is a noticeable shift. The sentiment dropped from **Positive** to **Negative** during the third execution loop, indicating increased difficulty with the task steps.
```

**Prompt:** 
```
Find where the agent realized it made a mistake.
```

**Response:** 
```
I found the following instance in your logs:

> "I made a mistake in the previous step. Let me try again."

This occurred during the **Correcting** phase of the execution.
```

## Capabilities

### Track error loops
See exactly how many times an agent repeats a failed step.

### Monitor progress states
Determine if the agent is currently correcting or proceeding through a task.

### Audit internal reasoning
Find specific parts of a log where the agent expresses internal thoughts.

### Measure sentiment shifts
Watch how the tone of logs changes during long-running executions.

## Use Cases

### Debugging infinite loops
An engineer sees an agent repeating a task and uses the MCP to confirm it is stuck in a correction cycle.

### Quality auditing
A developer reviews logs from a week of runs to find where high levels of uncertainty appear in the reasoning.

### Performance benchmarking
You compare two different prompts by checking which one has a lower self-correction rate.

### Agent stability testing
A researcher uses the tool to see if an agent maintains a consistent tone during long, complex tasks.

## Benefits

- Stop guessing if your agent is stuck by tracking error frequency.
- Identify loop patterns using detect_reflection to see where reasoning fails.
- Understand agent stability by measuring the rate of self-correction per loop.
- Detect performance drops by watching sentiment shifts with analyze_sentiment.
- Get clear status updates on whether an agent is proceeding or correcting.

## How It Works

The bottom line is you get measurable visibility into how your agent thinks and recovers from errors.

1. Connect your agent's execution log source to the MCP.
2. Point the tool at your recent history or specific log files.
3. Get a breakdown of self-correction rates and sentiment trends.

## Frequently Asked Questions

**How can I use Agent Self-Reflection & Sentiment Scanner to find errors?**
It scans your logs for specific markers like 'I made a mistake' or 'error'. This helps you pinpoint exactly where the agent recognized a failure.

**Can Agent Self-Reflection & Sentiment Scanner track agent progress?**
Yes, it determines if the current state is 'Correcting' or 'Proceeding' based on the markers found in your execution history.

**Does Agent Self-Reflection & Sentiment Scanner work with any logs?**
It works with any text-based execution logs from your AI client, making it easy to monitor different types of autonomous workflows.

**How does Agent Self-Reflection & Sentiment Scanner help with debugging?**
It quantifies how often an agent loops on errors, giving you a measurable way to see if your prompts are actually working.

**Can I monitor sentiment trends with Agent Self-Reflection & Sentiment Scanner?**
Yes, you can track how the emotional tone of the logs changes over time to identify when an agent is struggling.