# Humanloop MCP for AI Agents AI Agent Connect

> Humanloop lets you manage your entire prompt library directly from your AI agent. Instead of jumping between tabs to update versions, track logs, or push new prompts to production, you can handle it all through a natural conversation. It connects your organization's prompt engineering workflow to any MCP-compatible client, making LLMOps feel like a first-class feature of your development environment.

## Overview
- **Category:** developer-tools
- **Price:** Free
- **Endpoint:** https://edge.vinkius.com/vk_preview_SDhmK5J7LYd6XJCbZRuzCR6WIRna9REWi8m4RQ2M/ai-agent-connect
- **Tags:** prompt-engineering, llmops, prompt-management, ai-deployment, version-control

## Description

This Connector connects your Humanloop account to any AI agent to handle your prompt engineering workflow. Instead of jumping between tabs to update versions, track logs, or push new prompts to production, you can handle it all through a natural conversation. It turns prompt engineering from a manual chore into a first-class feature of your development environment. Because Vinkius makes it easy to connect these types of tools, you can actually treat your prompts like real code. You get a centralized source of truth where you can see every iteration, track how models are responding in real-time, and swap out configurations across different environments instantly. It's about moving faster and with more confidence, knowing that the prompt your agent is using is exactly the one you intended to deploy. You can check which versions are live in staging, audit the history of a specific prompt, or roll back a deployment in seconds. It removes the friction of manual configuration and lets you focus on the actual content of your prompts. Whether you're pushing a hotfix to production or just checking the latest logs for a new feature, this connection ensures your AI agent always has the most up-to-date information from your Humanloop library.

## Tools

### delete_prompt_version
Permanently remove an old or incorrect version of a prompt. This helps keep your library clean.

### deploy_prompt
Set a specific prompt version as the active one for staging or production. Use this to push updates live.

### get_prompt
Fetch the full configuration and details for a specific prompt ID. This lets you inspect the current settings.

### list_prompt_environments
See which prompt versions are currently live in every environment. Use this to check production status.

### list_prompts
Get a complete list of every prompt in your organization. Use this to see your whole library.

### log_to_prompt
Save a model's generation to your logs for later evaluation. This helps you track performance over time.

### remove_deployment
Take a specific prompt version offline from an environment. Use this to roll back changes quickly.

### update_monitoring
Turn monitoring evaluators on or off for specific prompts. This gives you control over your logs.

### update_prompt_version
Change the name or description of an existing prompt version. This keeps your metadata accurate.

### call_prompt_stream
Execute a prompt and get the response back in a live stream. This is useful for seeing model output as it happens.

### list_prompt_versions
View the full history of versions for a single prompt. This is great for comparing iterations.

### upsert_prompt
Create a new prompt or update an existing configuration. Use this to manage your prompt library.

## Prompt Examples

**Prompt:** 
```
What prompts do we have in our organization?
```

**Response:** 
```
I've retrieved your prompts. You have 3 active prompts:

* **customer-support-reply** (ID: pr_1)
* **code-refactor-bot** (ID: pr_2)
* **marketing-copy-gen** (ID: pr_3)
```

**Prompt:** 
```
Show me the versions for the 'Summarizer' prompt.
```

**Response:** 
```
Fetching versions for 'Summarizer'...

| Version | Name | Status |
| :--- | :--- | :--- |
| v_987 | Production_Final | Active |
| v_986 | Staging_Test | Pending |
| v_985 | Draft_01 | Archived |
```

**Prompt:** 
```
Deploy version v_987 of prompt pr_123 to production.
```

**Response:** 
```
Deployment successful. Version 'v_987' of prompt 'pr_123' is now live in the 'production' environment.
```

## Capabilities

### Push prompt versions to production
Set a specific prompt version as the active deployment for any environment like staging or production.

### Stream live model responses
Execute a prompt and receive the response back in a live stream directly to your agent.

### Audit organization prompts
Retrieve a complete list of every prompt in your organization for a quick overview.

### Track model logs
Record model generations to your logs for evaluation and performance tracking.

### Toggle monitoring settings
Activate or deactivate evaluators for monitoring logs within a specific prompt.

### View deployment status
List all environments and see which versions are currently deployed for each prompt.

## Use Cases

### The Production Hotfix
An engineer needs to swap a failing prompt. They ask the agent to deploy_prompt to production immediately to fix a bug.

### The Audit
A PM wants to see what's live. They ask the agent to list_prompt_environments and see the current status of all prompts.

### The Rapid Iteration
An AI dev wants to test 5 versions. They use list_prompt_versions to see the history and get_prompt to check the details.

### The Log Review
A dev wants to see why a prompt failed. They ask the agent to find the latest log_to_prompt entries for a specific ID.

## Benefits

- Deploy updates faster by using deploy_prompt to push versions to production without manual dashboard clicks.
- Keep your library clean by using delete_prompt_version to prune old iterations.
- Improve reliability by using list_prompt_environments to see exactly what is live everywhere.
- Speed up testing by using call_prompt_stream to see live responses as they generate.
- Better oversight by using log_to_prompt to keep a record of every model interaction.
- Dynamic monitoring by using update_monitoring to toggle evaluators as needed.

## How It Works

The bottom line is you get a unified, conversational interface for managing your entire LLMOps pipeline.

1. Connect your Humanloop API key to your AI client via Vinkius.
2. Ask your agent to find, update, or deploy specific prompts in your library.
3. Receive real-time status updates or streamed model responses in your chat.

## Frequently Asked Questions

**What is Humanloop MCP for?**
It is for managing your entire LLM prompt library. It allows you to version, deploy, and monitor prompts directly through your AI agent.

**Can I see what's live in production with Humanloop MCP?**
Yes, you can see the current deployment status for all your prompts. The Connector lets your agent list every environment and show which specific version is active right now.

**How do I update a prompt in Humanloop MCP?**
You can update a prompt by asking your agent to change the name or description of an existing version. It uses the upsert tool to keep your Humanloop library current.

**Can Humanloop MCP help with version control?**
Yes, it's built for that. You can list all versions of a specific prompt, retrieve history, and push new versions to specific environments like staging or production.

**Does Humanloop MCP support streaming responses?**
Yes, it does. Your agent can execute a prompt and stream the response back to you in real-time, which is much better for long-form content generation.

**How does Humanloop MCP handle logging?**
It allows your agent to record model generations as logs. You can then review these logs to evaluate how your prompts are performing in the wild.

**Can I turn off monitoring in Humanloop MCP?**
Yes, you can activate or deactivate evaluators for monitoring logs. This gives you full control over which prompts are being tracked for quality.

**Is Humanloop MCP good for production deployments?**
It's designed specifically for it. You can manage environment-specific configurations and deploy new versions to production with a single command from your AI client.

**Can I see the full version history of a specific prompt?**
Yes! Use the `list_prompt_versions` tool with the Prompt ID. It will return all historical versions, allowing you to track changes and metadata over time.

**How do I deploy a prompt version to a specific environment like production?**
You can use the `deploy_prompt` tool. Provide the Prompt ID and the Environment ID to set that specific version as the active deployment for that environment.

**Is it possible to record model outputs for later evaluation?**
Absolutely. Use the `log_to_prompt` tool to record a generation, including the prompt path, messages, and output, which can then be used for evaluation in Humanloop.