Choose the right way into rConfig
rConfig has several ways to connect it to other systems, and they’re easy to mix up. Some are for your code calling rConfig, some for rConfig calling out, and some run inside rConfig itself. This page puts them side by side so you can pick the right one before writing anything.
The options at a glance
| Option | Direction | Who builds it | Where it runs | Use it when |
|---|---|---|---|---|
| REST API | Your code calls rConfig | Your team | Anywhere | Your system needs to read from rConfig or change it |
| Webhooks | rConfig calls your code | Your team (the receiver) | Anywhere reachable from rConfig | Your system needs to react when something happens |
| MCP server | An AI agent calls rConfig | Nobody: it’s built in | Your agent | An engineer wants an agent to answer questions from rConfig data |
| API Collections | rConfig calls a platform’s API | Configured, not coded | Inside rConfig | You want to back up a cloud or controller platform that has a REST API |
| Script Integration Engine | rConfig runs your script | Your team | On the rConfig server | You need to collect a config from a device rConfig can’t reach by itself |
| Custom Commands | Code inside rConfig | rConfig Professional Services only | Inside the application | An integration needs deep access to rConfig’s internals |
| Custom Workflows | Code inside rConfig | rConfig Professional Services | Inside rConfig | Multi-step service automation for ISPs and MSPs, such as site provisioning |
Two questions that usually decide it
Who starts the conversation? If your system needs data or wants rConfig to do something, your code calls the REST API. If your system needs to know the moment something happens (a config changed, a device failed, a compliance check failed), rConfig calls you through a webhook. Many integrations use both: a webhook says something happened, and the receiver calls the API for the detail.
Is it a person with an AI agent, or a program? MCP is for agents acting on a plain-language request. It is read-only and needs no code. The REST API is for programs that know exactly which calls they make, and it can write as well as read. The tokens are separate: an MCP token only works on the MCP endpoint, and a REST API token can’t be used for MCP.
REST API compared with MCP
| REST API | MCP server | |
|---|---|---|
| Caller | Your code: scripts, pipelines, portals | An AI agent acting on a plain-language request |
| Access | Read and write, limited by token abilities | Read-only |
| Authentication | REST API token in the apitoken header |
MCP token as a bearer token |
| Scope | What the token owner’s role allows | Devices the token owner’s role can see |
An agent can also write code against the REST API for you. See developer tooling.
Easy to confuse
- API Collections is not the REST API. With API Collections, rConfig is the client: it calls another platform’s API and stores the response as a versioned config. The REST API is rConfig’s own API, which your code calls.
- Device Sync is not something you need to build. rConfig already keeps its inventory in step with NetBox, Nautobot, Infrahub, Zabbix and several monitoring tools. See the Device Sync overview before writing a sync of your own. The sync recipe is for sources Device Sync doesn’t cover.
- Chat notifications don’t need a webhook receiver. Slack, Teams, Discord and Google Chat are built-in webhook formats. Write a receiver only for systems that need the full JSON event.