“Shift for developers is more radical than we think.”
– Jeet patel, President and CPO, Cisco
AI operates a full throttle and leaves the awakening of radical changes for software developers. We understand the time when AI can write code, call tools and perform comprehensive workflows – all from one challenge. This shift has huge consequences. Read the blog of the jeet and learn more.
Radical shifts are on the horizon for more than just developers.
What about the impact of AI on network engineers?
In my previous blog I described the MCP – Model Context – and how AGR can finally talk about our language, understand our networks and take meaningful steps. Now I want to show you what happens when this conversation goes one step further: when the agent does not only understand what is broken but repaired it without being said how.
From detection to action: self -healing network
The solo network is not a hypothetical “co-i”. This is the agent AI manipulation with one of the most frustration of network operations: Drift configuration.
Let’s analyze the process of self -healing.
In this setting I put on Two MCP servers“Integrate with my example Splunk Enterprise and another integrated with my Meraki instrument panel.” What this special does is not just the integration of the tool – it is that the agent can make a autoomous decision across the stall tools by simple fast.
Server 1: Splunk MCP
On the SPLUNK MCP server, we start with such a natural language command:
- “” “”What are my indexes of joking? ”
- “” “”Search problems with changes in the last 4 hours or to change the status of Meraki_index. ”
- “” “”Show me all changes in network configuration and group them with the person who has produced them.“” “”
- “” “”Analyze network traffic formulas and identify any anomalies that could indicate safety threats.“” “”
The agent processes a request – under the hood – like this:
The SPLUNK MCP server uses SDK to ask the actual log data. Its task is to detect whether something around – such as a change in configuration – has occurred from what we expect, compare it to our source of truth and correct it.
Let’s try a self -healing with an agent we call “Network Pharaon”.
Here is a good challenge to start:
Call > I have to see what’s going on with my meaki network. You can show me the latest alerts of flushing, specifically in Meraki_index? I need to look into spar Where is the source Meraki network Only in the last 3 weeks.
You notice that I didn’t have to be specific about how to look. I just had to tell the agent what I was after.
Impressive, right? Here’s what Pharaoh did:
- Initial search attempt – Trying to search for Meraki_index with source = “Miaki network” but got a syntax error.
- FIXAT FIXAT Format – Edited Syntax searches to the correct format.
- Wide recognition – He searched the entire Meraki index to understand the structure of data and available sources.
- Data analysis – Found two main sources: “Home Network” and “Meraki Network” (configuration changes).
- Targeted extraction – Focus specifically on the source “Meraki Network” according to the desired.
- Warning analysis – Extracted key field: Types of alert, level, device, and details of changing from payload JSON.
All of this was in itself, including self -service and payment to get the result I required.
Server 2: Meraki MCP
The second MCP server is a place where it is “self -healing”.
After the Meraki MCP server receives a detected change (for example IP device Address changes), uses the Meraki instrument panel API to reverse these changes. No manual instructions, no pre -programmed chain of responsibility. The agent understood that the changed was a drift and took steps to restore alignment.
With me
- I didn’t have to write firmly encoded if it causes them. I just defined the tools and gave the agent context. The agent is a suitable tool, selected the right fun and fully caroom.
- I have defined tool decorators to make it available in my Meraki MCP-Do you are the most broken of the country that performs only one thing-add a list of my devices, update my device, so all that network engineers probably used and encoded.
This is what happens if you let the work manage the action and have an agent to do an orchestration. It’s simple, scalable and strong.
Now let’s look at the Howals agent for the MiaKi MCP network (which contained a real output).
We will first give gand the difference of what has changed
Call > This dude Kareem Iskander should not make any change in the network. Unacceptable! Can you show me the by -manufacturer, what has changed?
Once again, impressive! Note that the information is pulled out of the merger via the MCP SPLUNK server. Also manual, as our agent gave us suggestions on how to return changes. Once again, impressive! Note that the information is pulled out of the mold using the MCP SPLUNK server.
Also, notice how our agent gave us suggestions, how to automatically return changes using the available API endpoints in the MCP MCP! I did not have to be specific to which organization Meraki or this network network, I did not have to specify the type of device. Network Paraoh knew the hierarchy of Meraki dashboard and went through it!
Now it’s time to recover the net!
Call > NetP Let’s return the configuration to its original state for all the changes you have detected!
Why does it matter
This is not just a fun project. Deals with the actual bread point for all network engineers: Drift configuration!
Whether it is accidental changes, unauthorized adjustments or incorrectly settlement with the source of truth, the configuration drift leads to downtime, compliance and endless manual cleaning. AGRICT AGE offers a better model: Detect, understand and correct automatically.
I just took two steps and let the agent run with him:
- Define tool interfaces (SDK + MERAKI API)
- These tools register at MCP
This is the power of building Agency systems We already know at the top of the workflows.
What skills do you need?
Let’s keep it real. Here are the required skills:
- Coding with python
- Understanding of SDK and how to use them
- Automation and Programmability of Network with API
- MCP framework for access structure and tools
- Network skills
Where it all goes
Let’s get closer to a moment to better understand the overall image.
What is building-samolachical network using two MCP-not prototype agents. It is a practical view of the wider vision of Cisco.
At the beginning of the AI screen Anna Cisco laid the foundation for the agent era: modular agents working with our instruments, understanding our intention, and doing a caromaous action that fits directly into one agent who detects drift through merger.
Now imagine layering in a deep sip in the area of Cisco networks, a readable understanding of your Entitritire network, trained for years of expertise and telemetry at the CCIE level, and a number of predetermined agents prepared from the box.
Instead of simply reverse the incorrectly configured VLAN, the agent understands:
- Which app depends on the fact that the VLAN across the hybrid shit
- Whether a change introduced to a violation of segmentation or power regression
- How to solve the problem without critical business traffic
- How to update the source of truth to reflect any legitimate intention for change
This is where the theory becomes a reality:
- Ai canvas It gives us the environment and agents.
- Model Deep Network Cisco Deep It gives agents situational intelligence to deal with context.
- MCP It provides usoa prolonged (bring/create your own agent, which is a future function).
And this is what engineers need – not another platform, but an assistant who gets it; The one that can think like us works faster than we do and decides which we trust.
It’s time to sit on the driver’s seat
This is not one -time. It multiplies. Together, AI canvas, model Cisco Deep Network and MCP Network engineers In the seat of the driver of this new agent AI era. As he said, too, “The future comes quickly than you think.”
Stay in front of the curve and be part of the extraordinary.
For a fully functional code of this demo you can find on my Github restitization.
Unlock the future of technology using artificial intelligence training in Cisco U.
Explore AI learning and start building your skills today.
Read more from the AI Break series:
Sign up for Cisco U. | Join Cisco Learning Network today.
Learn with Cisco
X| Fibers Facebook | LinkedIn | Instagram| YouTube
Wears #Ciscou and#CiscocertJoin the conversation.
Share:
(tagstranslate) cisco u.