The rConfig MCP server: code meets configuration
Tap into the Model Context Protocol to generate scripts, automate workflows, and explore configurations securely using your AI coding assistant. It is the outbound half of our network AI maturity model.
AI for network reliability
Bridge the gap between human intent and machine execution with the power of the Model Context Protocol.
Configuration drift
AI-assisted detection of unintended changes across your fleet, with corrections proposed for human approval.
Automated scripts
Turn natural language into ready-to-run automation code for Python, Bash, or Ansible.
Open developer access
AI can browse, query, and take governed, permissioned actions on rConfig’s open-source foundation through the standardised MCP.
Private & secure
Absolutely no data leaves your environment. The MCP server runs locally within your infrastructure.
How MCP-AI works
Four stages, connect, engineer, explore, govern, each built to surface context without ever leaving the customer's rConfig instance.
Connect your AI tools
Seamlessly integrate with Claude Code, Cursor, and ChatGPT. The MCP standard allows these models to automatically discover rConfig's capabilities without complex manual configuration. Native Claude & Cursor integration · automatic tool discovery via MCP · multi-model orchestration support.
Intelligent engineering
Create backup scripts, diff API requests, and reporting pipelines in seconds. Just describe what you need, and watch the code appear, ready for review and secure execution. Precision script generation · real-time API diffing & testing · governed workflow automation.
Explore rConfig’s core
Your AI assistant can securely browse models, documentation, and configuration metadata, giving it the specific context it needs to provide accurate, network-aware answers. Secure metadata indexing · deep configuration exploration · live documentation grounding.
Enterprise governance
Full RBAC inheritance with a strict separation between read-safe and write-class operations. Read-safe is the default. Any write-class action is permissioned, human-approved, and logged for full audit compliance. Strict RBAC inheritance · read-safe by default, writes gated · full audit logging for AI sessions.
AI as a Skill, built into your network toolkit
Download rConfig as a Skill and your AI coding tools gain network-aware context and governed actions, without bespoke integration work.
A packaged capability, not just an endpoint
The MCP server is the live endpoint. The rConfig Skill is the packaged capability your AI tools load to learn how to drive rConfig, distinct from the server and portable across tools.
Loaded by your AI coding tools
Claude Code, Cursor, ChatGPT and other MCP-aware tools load the rConfig Skill to gain network-aware context and a governed set of actions, scoped by your RBAC.
Automatic discovery, no manual wiring
Through the MCP standard the tool discovers rConfig's capabilities for itself, so the developer gets network context without complex manual configuration.
Model Context Protocol, under the hood
rConfig exposes a curated “toolset” to AI models via MCP. These tools describe available API calls, schemas, and safe operations in a format LLMs natively understand.
- Describe API definitions & schemas
- Distinguish safe vs. unsafe operations
- Context-aware config inspection
- Tool discovery
- Code generation
- Review & run
- Audit logged
All executed inside the customer’s rConfig instance. No external data egress.
The workflow
From prompt to production in five reviewable, audit-logged steps.
Natural-language request
Ask your AI assistant to perform a task.
Tool discovery
AI identifies available rConfig MCP tools.
Code generation
AI constructs the necessary API calls or scripts.
Review & run
You validate the logic, then execute securely.
Audit logged
Full traceability of the AI’s actions.
Who uses MCP-AI
Same underlying protocol, three different jobs, each team gets configuration-aware automation phrased for their workflow.
System integrators
Build and maintain complex automations faster. Generate boilerplate integration code for third-party tools instantly.
MSPs & DevOps
Generate multi-tenant onboarding scripts and compliance reports tailored to specific customer environments safely.
Automation engineers
Turn plain-language directives into Python, Bash, PowerShell, or PHP pipelines without manual syntax lookup.
Frequently Asked Questions
Is my network data sent to external AI models?
No. The rConfig MCP server runs inside your own infrastructure. The AI model handles the reasoning, but data retrieval and tool execution happen locally, and you control exactly what context is exposed. Nothing about your network is shared unless you grant a tool that returns it.
Can the AI make changes to my devices?
Only under your control. Read-safe is the default. rConfig separates read-safe tools from write-class tools, so you can keep the AI in read-only mode for analysis, or grant scoped write-class tools where every change is permissioned by RBAC and approved by a human before it runs. Nothing writes to your network on its own.
What is the rConfig Skill, and how is it different from the MCP server?
The MCP server is the live endpoint your AI tools connect to. The rConfig Skill is the packaged capability those tools load to learn how to drive rConfig, portable and distinct from the server. The server is the connection; the Skill is the know-how the tools consume.
Which AI tools work with the rConfig MCP server?
Any client that speaks the Model Context Protocol, including Claude Code, Cursor, ChatGPT and other MCP-aware tools. Because discovery is handled by the standard, the tool finds rConfig's capabilities for itself without bespoke integration work.
How are write actions governed and approved?
Write-class tools are opt-in and scoped. A write is proposed by the AI, checked against your RBAC, and approved by a human before it runs, with the whole exchange logged for audit. The present posture is read-safe by default, and nothing writes autonomously.
Is access controlled by my existing RBAC?
Yes. The MCP server inherits rConfig's role-based access control, so an AI session can only reach what the connecting user is already permitted to see and do. There is no separate permission model to maintain, and every tool call is audited.
Do I need to write code to use the MCP integration?
No. You connect an MCP-aware tool and ask in plain language, and the rConfig Skill and the MCP standard handle the wiring. When you do want automation, the AI can generate Python, Bash, PowerShell or PHP for you to review and run, but getting started needs no code.
Do I need a special licence for MCP features?
MCP support is included in rConfig Pro and Enterprise editions. It builds on the existing API architecture, so your current RBAC and security policies apply automatically, with no separate AI add-on to license.
Bridge your AI with your Infrastructure
Deploy the rConfig MCP Server to give your AI assistants direct, governed access to your network data. Secure, scalable, and built for the enterprise.