About Obelinf
Obelinf MCP Server connects AI agents to Obelinf, a network and infrastructure documentation platform. It supports querying and updating infrastructure records through a hosted endpoint, with scoped read-only or full-access API keys and an audit trail for writes.
Key features
- 01Search across infrastructure records
- 02Manage sites, racks, and devices
- 03Track IPs, subnets, VLANs, and VRFs
- 04Manage cables, MAC addresses, and contacts
- 05Review field-level change history
Use cases
- 01Look up network inventory from an AI agent
- 02Update infrastructure records through MCP
- 03Inspect IP address and VLAN assignments
- 04Review attributed changes to infrastructure data
MCP support
Provides 52 tools over a hosted Streamable HTTP endpoint. ChatGPT connects with OAuth; other clients use scoped API keys for read-only or full access.
Configuration
Remote servers need no local install. Pick your client, then copy the config or install in one click.
https://api.obelinf.com/mcp?org={orgSlug}claude mcp add --transport http obelinf https://api.obelinf.com/mcp?org={orgSlug}{"mcpServers":{"obelinf":{"type":"http","url":"https://api.obelinf.com/mcp?org={orgSlug}"}}}Open in Cursor{"servers":{"obelinf":{"type":"http","url":"https://api.obelinf.com/mcp?org={orgSlug}"}}}Open in VS Code{"mcpServers":{"obelinf":{"type":"http","url":"https://api.obelinf.com/mcp?org={orgSlug}"}}}{"mcpServers":{"obelinf":{"serverUrl":"https://api.obelinf.com/mcp?org={orgSlug}"}}}Run commands in a terminal; paste JSON into the client’s MCP config and restart. Servers that need sign-in will prompt on first connect.
Tools
Details
From the project READMEObelinf MCP Server
Hosted Model Context Protocol server for Obelinf, the network and infrastructure documentation platform. Connect AI agents like Claude, Cursor, or Codex to your infrastructure records and let them query, answer, and update sites, racks, devices, IP address space, and cabling through one endpoint.
This repository does not contain the server's source code. It holds the MCP listing metadata, connection docs, the changelog for the MCP surface, and the public issue tracker.
- Website: https://obelinf.com
- API documentation: https://obelinf.com/docs/api/
- MCP feature overview: https://obelinf.com/features/ai-infrastructure-documentation/
Quick start
- Sign in at obelinf.com and open your organization. MCP is available on all plans, including the free Personal plan.
- Create an API key: Dashboard → API Keys → create a key and choose a scope:
Read OnlyorFull Access. Copy it when shown; it is only displayed once. - Add the server to your MCP client. Your endpoint is
https://api.obelinf.com/mcp. The organization is selected from your authenticated Obelinf account.
ChatGPT
ChatGPT uses OAuth. Add the MCP endpoint and complete the Obelinf approval flow when prompted. Google and GitHub remain the available Obelinf sign-in methods; ChatGPT receives a short-lived, organization-scoped MCP token after approval.
Cursor (plugin)
Install the Obelinf/mcp repository as a Cursor plugin from the Cursor directory. Set OBELINF_API_KEY once under Plugins → Configure and the server connects automatically. No manual JSON editing needed.
To add it manually instead, use the config below.
Claude Code
claude mcp add --transport http obelinf "https://api.obelinf.com/mcp" \
--header "Authorization: Bearer YOUR_API_KEY"
Claude Desktop, Cursor, and other JSON-configured clients
{
"mcpServers": {
"obelinf": {
"url": "https://api.obelinf.com/mcp",
"headers": { "Authorization": "Bearer YOUR_API_KEY" }
}
}
}
VS Code (.vscode/mcp.json)
{
"servers": {
"obelinf": {
"type": "http",
"url": "https://api.obelinf.com/mcp",
"headers": { "Authorization": "Bearer YOUR_API_KEY" }
}
}
}
Other MCP clients: point them at the URL above over Streamable HTTP and send the API key as an Authorization: Bearer header.
Endpoint details
| Property | Value |
|---|---|
| Endpoint | https://api.obelinf.com/mcp |
| Transport | MCP Streamable HTTP |
| Authentication | OAuth for ChatGPT; scoped API keys for other clients |
| Rate limit | 100 requests per minute |
| Availability | All plans |
| Tools | 52 |
Tools
| Group | Tools | Description |
|---|---|---|
| Search | search | Cross-resource full-text search across your inventory |
| Sites | list_sites, get_site, create_site, update_site, delete_site | Physical locations and data center sites |
| Racks | list_racks, get_rack, create_rack, update_rack, delete_rack | Rack inventory, units, and positions |
| Devices | list_devices, get_device, create_device, update_device, delete_device | Servers, switches, appliances, and VMs |
| Subnets | list_subnets, get_subnet, create_subnet, update_subnet, delete_subnet | IPAM subnets and prefixes |
| IP addresses | list_ip_addresses, get_ip_address, create_ip_address, update_ip_address, delete_ip_address | Assigned and free addresses |
| VLANs | list_vlans, get_vlan, create_vlan, update_vlan, delete_vlan | VLAN groups and IDs |
| VRFs | list_vrfs, get_vrf, create_vrf, update_vrf, delete_vrf | Virtual routing and forwarding instances |
| Cables | list_cables, get_cable, create_cable, update_cable, delete_cable | Cabling between ports and endpoints |
| MAC addresses | list_mac_addresses, get_mac_address, create_mac_address, update_mac_address, delete_mac_address | MAC inventory and assignments |
| Contacts | list_contacts, get_contact, create_contact, update_contact, delete_contact | People and vendors tied to infrastructure |
| Changelog | list_changelog | Audit trail of every change, with actor and field-level diffs |
Permissions
- Read Only keys can use all read and search tools. Write tools return an error.
- Full Access keys can read and write. Every write is recorded in the organization changelog with attribution and field-level diffs, so agent edits are as reviewable as manual ones.
Start agents on a Read Only key, review their diffs, then promote to Full Access deliberately.
Support
- Bug reports and feature requests for the MCP surface: open an issue
- Plans, pricing, and enterprise agreements: Contact sales
Comments
New to MCP?
From what the protocol is to installing your first server, these answers cover most first questions.
An open protocol that defines how AI applications connect to external tools and data in a uniform way. Clients talk to the model; MCP servers provide tools, resources and prompt templates. They exchange standard messages, so one server can be reused by many clients.