MCP server setup
Connect your AI assistant to Aplas using the MCP server
Prerequisites
You will need an Aplas API key. To generate one:
- Log in to Aplas and open Config → API.
- Click Create API Key, choose its role, and copy the generated key.
The role decides what an assistant connected with that key may do:
| Role | What the assistant can do |
|---|---|
| ReadOnly | Search, count, and read assets and their relationships. Every tool that changes data is refused. |
| ReadWrite | The above, plus creating and updating assets and grouping them. |
A ReadOnly key is the right default. It lets an assistant answer questions about your estate with no possibility of altering it — useful when you are trying MCP out, or connecting a client you do not fully control.
Keep your API key secure
Your API key provides access to your organization's data. Do not share it publicly or commit it to version control.
Regional endpoint
Aplas is hosted in three regions and the MCP server lives at the same regional host as the REST API, under /mcp/v1:
| Region | MCP URL |
|---|---|
| Australia | https://api.au.aplas.com/mcp/v1 |
| Europe | https://api.eu.aplas.com/mcp/v1 |
| United States | https://api.us.aplas.com/mcp/v1 |
Use the URL that matches your organization's region. To find yours, open the Config dashboard in Aplas — the Data residency card shows your region and API endpoint. The examples below use the Australia URL; substitute your own.
Claude Code
Run the following command in your terminal:
claude mcp add --transport http aplas https://api.au.aplas.com/mcp/v1 --header "Authorization: Bearer YOUR_API_KEY"Cursor
- Open Settings > MCP.
- Click Add new MCP server.
- Add the following configuration:
{
"mcpServers": {
"aplas": {
"url": "https://api.au.aplas.com/mcp/v1",
"headers": {
"Authorization": "Bearer YOUR_API_KEY"
}
}
}
}VS Code (GitHub Copilot)
Create or update the .vscode/mcp.json file in your project:
{
"servers": {
"aplas": {
"type": "http",
"url": "https://api.au.aplas.com/mcp/v1",
"headers": {
"Authorization": "Bearer YOUR_API_KEY"
}
}
}
}Claude Desktop, claude.ai and the mobile apps
These connect Aplas as a custom connector. Anthropic's infrastructure calls our server rather than your device, which works because our MCP endpoints are on the public internet.
Request header authentication is in beta
Authenticating a custom connector with an API key uses Claude's Request headers, which Anthropic is rolling out gradually. If you do not see that section when adding the connector, contact Anthropic for access — or use Claude Code, Cursor or VS Code, which need no beta feature.
-
Open the add-connector dialog:
- Free, Pro and Max — Customize → Connectors → Add custom connector
- Team and Enterprise — an owner adds it under Organization settings → Connectors → Add → Custom (choose Web if asked for a type); members then connect it from Customize → Connectors
-
Enter the MCP URL for your region, for example
https://api.au.aplas.com/mcp/v1 -
Open Request headers and choose the
authorizationheader -
Enter the value including the scheme:
Bearer YOUR_API_KEYClaude sends the value exactly as typed and adds no prefix —
YOUR_API_KEYon its own is rejected -
Click Add, then enable Aplas from the + menu in a conversation
Because the credential is stored once for the connector rather than per person, everyone using it shares that key's role. A ReadOnly key is the safer default for a shared connector.
Verifying the connection
Once configured, ask your assistant:
List my Aplas workspaces
A working connection returns your workspaces, each with a short id — its slug.
That request is also the first step of every session. Aplas tools are scoped to one workspace, so
the assistant has to know which one you mean before it can search or write. A good client calls
list_workspaces on its own and then passes the slug to everything else; if yours does not, naming
the workspace in your question ("in the enterprise workspace, how many applications are Retired?")
is enough.
Protocol version
The server implements MCP protocol 2025-11-25, over streamable HTTP with no session state — each call is a self-contained POST carrying its own credential. Any client supporting that revision works; the three above all do.
Troubleshooting
Common issues
- 401 Unauthorized — Your API key is invalid or missing. Verify the key in Config → API and check that the
Authorization: Bearer <key>header is set correctly. - Wrong region — API keys are scoped to a single region. If you're hitting the right URL but still getting auth errors, confirm the key was issued in the same region as the URL.
- Connection timeout — Ensure your network allows outbound HTTPS connections to your regional
api.<region>.aplas.comhost. - No tools available — Restart your AI assistant client after adding the MCP server configuration.
- A write was refused — The key is ReadOnly. The refusal names the role the tool required; create a ReadWrite key in Config → API if the assistant should be able to change data.
- The server reports it is unavailable for your organization — Your organization is on the earlier (v1) data model, which this server does not serve; the message says so and points at the REST API that does. See the API overview for which version applies to you.