Orchestrating Real-world Agentic Workflows

In the world of music, timing is everything… and the same goes for operations. At a company focused on giving artists access to the world's most talented session musicians, their community of musicians, producers, and creators thrives on real-time collaboration. But with that much creative energy flowing through their Slack channels, it was becoming harder to catch every important request, respond promptly, and make sure the right team members were looped in.

That’s where this collaboration began: building a Gentoro MCP Tool that could listen to the right Slack channels, detect when action was needed, and instantly trigger the right workflow.

We call it a “Tool” because that’s exactly what it is: a highly capable MCP module that can be embedded into any AI agent, workflow automation system, or even run standalone via a webhook. In this specific case, the MCP Tool is triggered by Slack messages.

However, this project went beyond simply about automating responses. We set out to create a reliable, extendable MCP Tool that could streamline internal processes while respecting the nuance of human communication.

The Objective: An Agentic Tool That Gets It

The mission was clear:

Create a Gentoro MCP Tool that processes Slack messages, detects scenarios defined in a comprehensive runbook, and executes the right actions.

In other words, the Tool would read Slack messages in real time, detect when help was needed, and trigger the exact right workflow, whether that meant notifying a department, creating a ticket, or even sending an SMS depending on the urgency and priority of the issues.

To make this possible, the agentic tool needed to integrate with our essential tools:

• Slack → The main input and output channel. All the magic starts and ends here.
  
  
• Hubspot → For creating Customer Success tickets.
  
  
• Notion → Their central repository for issue and feature tracking.

The Architecture: Stable, Scalable, and Upgrade-Friendly

We deployed the Gentoro MCP runtime inside our client’s AWS infrastructure, with a design that prioritizes both stability and scalability.

When a musician posts a message in an observed Slack channel, Slack sends the message payload to an AWS Lambda function via an incoming webhook. The Lambda function acts as the gateway, preprocessing the message and extracting corresponding metadata, then forwarding it to the Gentoro AI agent tool in the runtime. The AI agent (in this case, a custom-built message classifier MCP Tool) analyzes the content acknowledging the musician’s message automatically, determines the correct scenario (e.g., technical issue, feature request, customer support), and executes the relevant workflow by integrating with corresponding tools in Hubspot, Notion, or sending Slack notifications. This architecture keeps Slack responsive, isolates processing logic in the MCP Tools, and ensures the workflow remains modular and easily extendable.

‍

We also provide an upgrade process desiged for zero downtime for the client: 

• Gentoro delivers new versions as Docker images.
  
  
• The client spins up a new AWS EC2 instance for each upgrade.
  
  
• AWS Lambda rotates traffic gradually (X% at a time) to the new version.
  
  
• Old instances are kept for at least a week to enable instant rollback.

This setup means Gentoro can ship updates quickly without risking downtime, which is something that our client’s operations team and musician community greatly appreciates.

Lessons Learned: Designing a Tool That Mimicks Human Interaction

The real challenge wasn’t integrating APIs, but rather making the MCP Tool feel like a helpful colleague rather than a robotic script. Slack conversations are full of nuance, casual language, and emojis. So, the Tool needed to capture these nuances and interact with the musician and client team in a natural way. 

We solved this by designing scenario-specific detection logic backed by a clear runbook, ensuring that the MCP Tool acts consistently while still leaving room for human handoff when needed.

What’s Next: From Slack Processor to Multi-Channel Operator

While this MCP Tool currently processes Slack messages, it can easily be extended to other channels, agents, or orchestration systems. Future plans include adding proactive reporting, richer context-awareness, and even predictive capabilities that could preemptively identify potential issues.

For now, the Gentoro-powered MCP Tool is helping our client’s teams respond faster, reduce manual follow-up, and keep the music flowing without operational noise getting in the way.

And just like in music, the best performances happen when everyone knows their cue.. now, Gentoro MCP toolbox does too.

This project was also featured on The Context livestream where we walked through the architecture and design decisions in detail. If you’d like to see the MCP Tool in action and hear more about how Gentoro makes it possible to build MCP Tools that feel as natural as a teammate, check out the recording: 

‍