Config Analysis & Explain
Chat about any device config, or select lines and ask for an explanation, issues, improvements, or a security review. Learn more
What is the AI Center in rConfig V8?
What is the AI Center in rConfig V8?
Section titled “What is the AI Center in rConfig V8?”The AI Center puts an AI assistant next to the configs, diffs, and compliance results you already have in rConfig, using a model you choose. It explains; it never changes anything on your network.
From V8.3.0, the AI Center helps with moments like these:
- “What does this config actually do?” You’ve inherited a firewall with 400 lines of ACLs and no documentation. Open it, click Ask AI, and get a walk-through in plain English.
- “What changed overnight, and is it why the branch is down?” Click Analyze with AI on the change and get a summary, the operational impact, and a risk rating before you’ve read the diff.
- “Why did this device fail the audit?” Click Narrate with AI on the compliance result and paste the explanation into your audit response.
- “Which devices still have SNMP community public?” Ask Claude Code or Cursor, connected to rConfig over MCP, and get the list from every stored config at once.
How it works
Section titled “How it works”flowchart LR
U["You click<br/>Ask AI / Analyze / Narrate"] --> R["rConfig<br/>adds context and<br/>checks permissions"]
R --> P["Your AI provider<br/>cloud API or local model"]
P --> R
R --> A["Answer shown<br/>in rConfig"]
R -.-> L[("Audit trail<br/>metadata only")]
- You ask from a button next to your data, so the AI already knows which config, diff, or result you mean.
- rConfig adds context: the config or diff, plus networking knowledge and any conventions you’ve taught it. It only uses data you’re allowed to see.
- Your chosen provider answers. rConfig doesn’t host a model. Requests go from your rConfig server to the provider you’ve configured, which can be a model on your own network.
- rConfig records the request in the audit trail: who, which capability, which model, and how many tokens. Never the prompt or answer text.
What’s in the AI Center
Section titled “What’s in the AI Center”Click AI Center in the left sidebar. The tabs across the top are:
| Tab | What it’s for | Who sees it |
|---|---|---|
| Overview | Provider status, which capabilities are on, and links to MCP, Skills, and API Docs | Everyone with AI Center access |
| AI Settings | Add, test, and activate an AI provider | Admins |
| AI Configuration | Turn each capability on or off | Admins |
| MCP Server | Connect Claude Code, Cursor, and other agents; manage MCP tokens | Everyone with AI Center access; creating tokens needs aicenter.mcptoken (details) |
| Skills | Download the rConfig Skill package for agents | Everyone with AI Center access |
| API Docs | Download REST API docs written for AI agents | Everyone with AI Center access |
| Privacy | What each capability sends, and where your data goes | Everyone with AI Center access |
Inside rConfig: four capabilities
Section titled “Inside rConfig: four capabilities”Drift Detection & Narration
Turn a config diff into a summary with operational impact and risk. Learn more
Policy & Compliance Narration
Explain compliance results in plain English, and draft new policy rules. Learn more
Optimization Assistant
Ask how to tune PHP, the database, and Horizon, answered from rConfig’s docs. Learn more
Outside rConfig: your own agents
Section titled “Outside rConfig: your own agents”- MCP Server: a live, read-only connection that lets agents look up devices, configs, diffs, and compliance status.
- Skills: a package that teaches your agent how to use rConfig’s tools well.
- API Docs: REST API docs for agents that write integrations and scripts.
Read-only, by design
Section titled “Read-only, by design”No AI feature in rConfig pushes a change to a device. The AI behind drift and compliance narration is instructed never to propose pushing changes, and no AI code path writes to a device or a stored config. The one thing AI can write is a draft policy definition, which lands in the editor and is only saved when a user clicks save. The MCP server has no write tools at all.
Bring your own model
Section titled “Bring your own model”Choose from Anthropic, OpenAI, Google Gemini, Azure OpenAI, Ollama, or any OpenAI-compatible server. With Ollama or a self-hosted server on your network, configs never leave your environment. See Set up an AI provider.