NetPilot

Custom AI Agents for Network Engineering

NetPilot builds your custom AI agent, the harness and the tools, tests them on real network operating systems, and delivers end to end.

Watch Demo
app.netpilot.io

See It in Action

Watch the agent design, build and validate a multi-vendor network on real network OSes. The same lab proves every agent we deliver.

Built and proven on real multi-vendor network OSes

17+
network OSes and growing, on real CLIs
5 steps
from scoping call to delivered agent
4 ways
to deploy, from NetPilot cloud to your own environment
1
guarantee: a money-back success milestone

NetPilot is a containerlab contributor and integrator. The NetPilot team built the mikrotik_chr kind in containerlab v0.82.0. See our upstream work.

NetPilot delivers the agent

Agent platforms sell a builder and leave the build to you. NetPilot owns the build and the outcome, with a guarantee.

We own the build

NetPilot builds the agent, the harness and the tools, and delivers the working result.

Agent platforms

You get a builder, and the build is your team's project.

Proven before delivery

Every agent is tested on real network operating systems in a lab that mirrors your network.

Agent platforms

Testing depends on what the platform and your team put in place.

Any vendor, any system

17+ network OSes and growing, plus any system with an API or an MCP server.

Agent platforms

Strongest on the vendor's own gear and data.

A guaranteed outcome

A success milestone agreed up front. Money back if it is not delivered.

Agent platforms

A license fee. The outcome is yours to deliver.

The offer, defined

What is a custom AI agent for network engineering?

A custom AI agent for network engineering is an agent built for one workflow your team runs: its own instructions, the tools it needs to reach your systems, and tests that prove it does the job. Name any network engineering agent. NetPilot builds the agent, the harness and the tools, tests it on real network operating systems, and delivers the working result.

Agent Studio is where every agent lives. Build your own, or have NetPilot build and deliver one for you. Delivered agents work in the lab and on your live network, with the approval steps your team sets.

Built and delivered end to endTested on real network OSesLab and live networkMoney-back success milestone

The agent, the harness and the tools

A working agent is three things. NetPilot builds all three and tests them together.

The agent

The instructions that tell the model how your team does this job: the checks to run, the order to run them in, and what to hand back.

The harness

The agent harness is everything around the model that keeps it on task: the steps it may take, the approvals it waits for, and the tests that prove each change.

The tools

How the agent reaches your systems: your inventory, your config repos, your ticketing and your devices, over API, MCP or SSH.

Network engineering agents we build

Each one is built for your workflow, your vendors and your systems. The list is examples, not a menu.

Pre-change validation

An agent that builds a lab copy of the affected segment, applies the change, compares routing and reachability before and after, and writes the change report.

Network research and testing

An agent that stands up multi-vendor test topologies, injects failures and impairment, and records how each vendor's implementation behaves.

Peering and interconnection

A BGP peering agent that checks candidates against public routing and interconnection data, drafts the session config for your network operating system, and verifies it in a lab before turn-up.

Provisioning and service turn-up

An agent that turns a service request into per-vendor configs, checks them against your standards, and prepares the change for approval.

Troubleshooting and incident triage

An agent that reads the alert, pulls device state with show commands, narrows the cause, and hands your engineer a tested fix.

Compliance and configuration audit

An agent that compares running configs to your golden standards, lists every drift by device and line, and drafts the remediation.

Migration and vendor swap

An agent that translates configs from one vendor to another, rebuilds the design in a lab, and proves feature parity before cutover.

Your workflow

If it is network engineering work, we build the agent for it. Bring the runbook your team repeats most.

See the lab workflows in depth: network change validation, network research lab and network migration lab.

Custom AI agent development services, in five steps

From a scoping call to a tested agent your team runs. NetPilot does the build.

  1. 1

    Scope the workflow

    A 30-minute scoping call with the founder. You walk through the workflow, the systems it touches and the result you need, and we agree a success milestone.

  2. 2

    Build the agent

    We write the agent's instructions and build the harness and the tools it needs for your inventory, your config repos, your ticketing and your devices.

  3. 3

    Test it on real network operating systems

    The agent runs in a lab that mirrors your network, on real multi-vendor network OSes, until it does the job the way your engineers would.

  4. 4

    Deliver it in Agent Studio

    The agent lands in your team's workspace. Your engineers start a chat with it, see every step it takes, and extend it when the workflow changes.

  5. 5

    Stand behind it

    If the agreed milestone is not delivered, you get your money back for that milestone.

NetPilot Agent Studio: where your agents live

One place for the agents you build and the agents NetPilot delivers, each proven on real network operating systems.

Agent Studio is where every agent lives

Each agent has its own instructions, model and tools. Build your own, or have NetPilot build and deliver one for you. Your engineers pick an agent and start a chat.

app.netpilot.io

Tested on real network operating systems

Before delivery, every agent runs against a lab that mirrors your network on real Cisco, Juniper, Arista, Nokia and other network OSes. The lab is the proof.

app.netpilot.io

Your engineers verify every step

The agent shows its work and the real CLI is always one click away. Your engineers stay in command: they set the intent, check the result and approve the change.

app.netpilot.io

Built and delivered, or build it yourself?

Four ways to get a network engineering agent, and what each one leaves to your team.

Primary use case
NetPilot
Custom agents for network engineering, change validation, rapid labs
Build it yourself on a framework
Teams with agent engineers on staff
Agent platform for networks
Monitoring-first network operations
General AI agent firm
General business workflows
Who builds the agent
NetPilot
NetPilot builds it and delivers it working
Build it yourself on a framework
Your engineers, on top of their day job
Agent platform for networks
Your engineers, with the platform's builder
General AI agent firm
The firm builds it
Tested on real network OSes before delivery
NetPilot
Yes, in a lab on 17+ network OSes
Build it yourself on a framework
Only if you build the lab too
Agent platform for networks
Depends on the platform's test tools
General AI agent firm
No network lab
Network depth
NetPilot
Built by network engineers, multi-vendor
Build it yourself on a framework
As deep as your team
Agent platform for networks
Deep on the vendor's own gear and data
General AI agent firm
Generalists, no vendor CLI experience
Always-on monitoring and telemetry
NetPilot
Reads from the monitoring you already run
Build it yourself on a framework
Your existing tools
Agent platform for networks
Built in, the platform's core strength
General AI agent firm
Out of scope
Control of the agent's code
NetPilot
Yours to run and extend in Agent Studio
Build it yourself on a framework
Full control
Agent platform for networks
Within the platform's limits
General AI agent firm
Varies by contract
Delivery guarantee
NetPilot
Money-back success milestone
Build it yourself on a framework
None, it is your project
Agent platform for networks
A license, the build is yours
General AI agent firm
Varies by contract
Where the agents run
NetPilot
NetPilot cloud or your own environment
Build it yourself on a framework
Wherever you host them
Agent platform for networks
Inside the vendor's platform
General AI agent firm
Wherever you host them

Bottom line

Pick an agent platform when you need:

  • •Your team wants its agents inside the monitoring platform it already runs
  • •Always-on telemetry and alerting is the main thing you are buying

Pick NetPilot when you need:

  • •You need a working agent and have no one free to build it
  • •The workflow spans several vendors and systems
  • •You want the agent proven in a lab before it acts on your network

Verdict:An agent platform suits a team that wants monitoring first and will build its agents there. NetPilot suits a team that wants the result: a tested agent, delivered and guaranteed, or its own agents built in Agent Studio.

The full comparison, with sources: build, buy, or have it delivered.

Deploy on your terms

The same platform and the same agents, wherever they have to run.

NetPilot cloud

Runs in the browser, with a dedicated VM per lab and a running multi-vendor topology in about 2 minutes.

Dedicated environment

Isolated, dedicated resources for your team, scoped to your workloads.

Self-hosted / on-prem

NetPilot and your agents in your own data center or private cloud, behind your perimeter.

Air-gapped AI

Fully offline on a local LLM (Ollama, vLLM, Foundry Local), with no cloud and no phone-home.

Your choice of AI provider per deployment: Claude, OpenAI, Gemini, AWS Bedrock, or a fully local LLM (Ollama, vLLM, Foundry Local) for air-gapped sites.

Security & data handling

The controls a security reviewer looks for, and the deployment options that keep your data where you need it.

Approval steps you set

A delivered agent acts on your network through the approval steps your team defines.

SSO

Single sign-on for your team.

Audit logs

Who deployed what, and who ran which CLI command.

Isolated environments

A dedicated, isolated environment per lab, with no shared tenancy.

Data residency

Keep data in your own environment, with your choice of AI provider, including a fully local LLM.

BYOI, no image distribution

We never distribute commercial vendor images. Bring your own licensed network OS images or sanitized configs.

Built for your security requirements. Tell us your compliance and data-handling needs on the scoping call, and we scope the deployment to meet them.

Two ways to work with our team

Both start with a 30-minute scoping call with the founder. Signature connects a system. Enterprise delivers an agent that owns a workflow.

Signature

Hands-on delivery, outcome guaranteed.

$500/ month

By application. Limited spots.

  • Unlimited AI and lab usage, fair use, for your agreed scope
  • Custom device builds: Cisco IOS-XE, Juniper vMX, NVIDIA Cumulus, or any containerlab-compatible image, built for you
  • Agent Studio: build and share your own agents with your team
  • Two-way MCP: your agents drive NetPilot's MCP server
  • Custom integrations: we connect a system you run, such as NetBox, Git or ServiceNow
  • Hands-on delivery with our team
  • Money-back success milestone
  • Direct support

Enterprise

Any network engineering agent, built and delivered for you.

Custom

Scoped to your workflow, deployment and AI provider.

  • Everything in Signature
  • Custom agent workflow delivery: we build the agent, the harness and the tools, test it, and deliver the working result
  • End-to-end delivery, backed by the delivery guarantee
  • Delivered in Agent Studio, where your team runs and extends the agents
  • On-prem, dedicated or air-gapped deployment
  • Your choice of AI provider (Claude, OpenAI, Gemini, Bedrock, or a local LLM)
  • Two-way MCP in your environment
  • SSO and audit logs

See the full plan comparison on the pricing page.

Use cases

Built for every team

One AI-native network lab, tailored to how each team works. Pick yours.

Network

Mirror production and validate BGP and ACL changes on real CLIs before they ship.

Explore Network change validation

Mirror, snapshot, validate

Pre/post snapshot diffs capture routing tables, neighbor state, and ACL behavior before deployment.

Multi-vendor in one sandbox

Cisco, Juniper, Arista, Nokia, Palo Alto, Fortinet, Aruba, Dell, MikroTik, and SONiC in one lab. The agent writes the per-vendor syntax.

~2 minutes, not days

About two minutes from a plain-English prompt to a working multi-vendor mirror on real NOS.

Rehearse the failure

Shut a backbone link or kill a BGP peer and verify reconvergence before touching production.

Custom AI agents for network engineering: FAQ

The questions enterprise network teams ask before they hand a workflow to an agent

A custom AI agent for network engineering is an agent built for one workflow your team runs, such as pre-change validation, peering, provisioning or troubleshooting. It has its own instructions, the tools it needs to reach your systems, and tests that prove it does the job. NetPilot builds the agent, the harness and the tools, tests it on real network operating systems, and delivers the working result in Agent Studio.
Any agent for network engineering work. Common requests are pre-change validation, network research and testing, peering and interconnection, provisioning and service turn-up, troubleshooting and incident triage, compliance and configuration audit, and migration or vendor swap. The list is examples, not a menu. If your team repeats a runbook, NetPilot builds the agent for it.
It depends on the workflow. NetPilot scopes each agent on a 30-minute call and prices Enterprise delivery for that scope, with a money-back success milestone. Signature, for hands-on delivery and custom integrations, is $500 a month. Building your own agents in Agent Studio is part of the paid plans listed on www.netpilot.io/pricing.
Yes. Agent Studio lets you build an agent with your own instructions, model and a chosen set of capabilities, and share it with your team. Building your own suits workflows your engineers can describe in a prompt. Delivery by NetPilot suits workflows that need custom tools, testing against your network and a result you can hold us to.
Build it yourself when you have engineers who know both networking and agent frameworks, and time to spare. Buy an agent platform when always-on monitoring is the main thing you need and you want agents inside that platform. Have it delivered when you need a working agent for a real workflow and nobody is free to build it. NetPilot delivers: we build the agent, the harness and the tools, test it on real network operating systems, and back it with a money-back success milestone.
Every agent runs in a lab that mirrors your network, on real network operating systems, before it is delivered. The lab is built from your topology and configs, the agent does the job end to end, and your engineers can SSH into any device to check the result. An agent is delivered once it passes the success milestone agreed on the scoping call.
An agent harness is everything around the model that keeps an agent on task: the steps it may take, the tools it can call, the approvals it waits for, and the tests that prove each change. For a network engineering agent, the harness is what stops a wrong command from reaching a device. NetPilot builds the harness with the agent and tests both on real network operating systems.
Yes. Every agent is first proven in a lab on real network operating systems. Once it passes, it works on your live network, read-only or with changes, through the approval steps your team defines. Lab workflows stay a primary use: pre-change validation, network research and testing all run on a lab copy of your network.
Yes, on the Enterprise plan. Deploy in your own data center or private cloud, or as air-gapped AI on your own LAN. For air-gapped installs the agent runs on a local LLM (Ollama, vLLM or Microsoft Foundry Local) with no cloud and no phone-home at runtime. For connected on-prem you can point the agent at Claude, OpenAI, Gemini or AWS Bedrock.
A custom integration connects one system. NetPilot connects to the network systems you already run, over API or MCP, and our team builds each integration for your workflow. NetBox, Nautobot, Git and ServiceNow are a few examples, and if it has an API, NetPilot connects to it. A custom agent owns a whole workflow and uses those integrations as its tools. Details: www.netpilot.io/integrations.
Agentic NetOps is the analyst term for AI agents that carry out network operations work under human control. NetPilot covers that work as delivered agents: each one is built for a workflow, tested in a lab, and runs with the approval steps your team sets. NetPilot is not a monitoring or telemetry platform. A delivered agent reads from the monitoring, inventory and ticketing systems you already run.
Enterprise AI agent platforms such as Microsoft Copilot Studio or ServiceNow AI Agent Studio give you a general builder for business workflows. They have no network lab and no knowledge of a routing protocol or a vendor CLI, and the build stays with your team. NetPilot is built for network engineering: the agent is written by network engineers, tested on 17+ real network operating systems, and delivered working.
Network automation services usually deliver scripts, playbooks or a platform rollout, billed for the time spent. NetPilot delivers an agent that does the workflow: it reads the request, works across your vendors and systems, checks its own result in a lab, and waits for your approval. You pay for a scoped outcome with a money-back success milestone.
17+ network operating systems and growing, on real CLIs. Nokia SR Linux, FRR and Linux are built in. Cisco IOL, Cisco XRd (IOS XR), Cisco NX-OS, Juniper cRPD, Juniper vJunos (router and switch), Arista cEOS, Palo Alto PAN-OS, Fortinet FortiGate, Aruba AOS-CX, Dell SmartFabric OS10, and MikroTik RouterOS (CHR) run with your own images (BYOI). MikroTik RouterOS (CHR) and SONiC (community VS) are preloaded on every lab VM, replaceable with your own versions. Custom images such as Cisco IOS-XE or Juniper vMX are built for you. NetPilot never distributes vendor images.
On the scoping call we agree a concrete success milestone with you, such as a working agent for a named workflow or a validated change. The work is hands-on and the milestone is defined in advance. If the agreed outcome is not delivered, you get your money back for that milestone. The guarantee is part of Signature and carries over to Enterprise.
Your team runs it. The agent is delivered in Agent Studio, where your engineers pick it, start a chat, and see every step it takes. They can change its instructions and extend it when the workflow changes, and NetPilot's team supports it under your plan.
Signature ($500/mo, by application) is hands-on delivery on NetPilot cloud: unlimited AI and lab usage for your agreed scope, custom device builds, two-way MCP, custom integrations with a system you run, Agent Studio for your team, direct support and a money-back success milestone. Enterprise (custom pricing) includes everything in Signature and adds custom agent workflow delivery end to end, plus deployment in your own environment with your choice of AI provider. Signature connects a system. Enterprise delivers an agent that owns a workflow.
Yes, and it is a primary use case. Describe a network in plain English and the agent designs the topology, writes the per-vendor configs, and deploys real network OS containers in about 2 minutes. You SSH into any device to verify the result. Delivered agents use the same lab to prove their work before it reaches your network.
SSO, audit logs (who deployed what and who ran which CLI command), dedicated and isolated environments per lab, and data residency through deployment in your own environment with your own choice of AI provider. We never distribute commercial vendor images, so you bring your own licensed network OS images or sanitized configs. We scope your specific compliance and data-handling requirements on the call and do not imply certifications we do not hold.
The scoping call is a 30-minute session with the founder. You walk through the workflow you want an agent for, the vendors and systems it touches, and the result you need. We define the agreed scope, name the custom tools and integrations to build, and set the success milestone. It is a working session, and you leave with a concrete build plan.

Explore the platform

Build, Buy, or Have It Delivered

Five ways to get a network engineering agent, compared on who builds it, who tests it and who owns the result.

Agent Studio Announcement

Build your own network agent, or have NetPilot build and deliver one.

Build an Agent in Agent Studio

The step-by-step guide to a custom AI agent for network engineering.

The Network Engineering Agent

Agents build the network. Engineers run the agents.

Integrations

Custom integrations with the systems you run. NetBox, Nautobot, Git and ServiceNow are a few examples.

Network Change Validation

Rehearse BGP, ACL and routing changes on a runnable copy of your network.

Network Research Lab

Reproduce cross-vendor behavior and inject failures for protocol R&D.

Network Migration Lab

Rehearse a vendor swap on a runnable copy of your network and prove parity before cutover.

Network Testing Lab

Multi-vendor test topologies on real network OSes, built from a prompt.

Tell us the agent you need

A 30-minute scoping call with the founder. We map the workflow, agree the success milestone, and tell you plainly what we will deliver.

See how the plans compare on the pricing page.