Connect Augure, Ollama & others
Connect Augure, an Ollama model, or any other MCP client to AssetLab with a server URL and an Authorization header.
Any client that speaks the Model Context Protocol can reach AssetLab. You give it the server address and send your API key as a header on every request.
Step 0 - create an API key
In AssetLab, an Administrator goes to Settings → API Keys → New key, names it for the client ("Augure - Maria"), picks scopes (start read-only; see Tools & scopes) and copies the al_live_... value - it's shown once.
The configuration
Most clients accept this shape. Replace al_live_... with your key:
{
"mcpServers": {
"assetlab": {
"url": "https://mcp.assetlab.ca/",
"headers": {
"Authorization": "Bearer al_live_..."
},
"transport": "http"
}
}
}
If your client asks for the fields one at a time: transport Streamable HTTP, URL https://mcp.assetlab.ca, header Authorization with the value Bearer al_live_.... A client that supports OAuth can skip the header - it opens the AssetLab authorization page, where you paste the key, the same flow as Claude.
Settings → AI Assistants shows the same configuration with a Copy button under each client.
Augure
Augure is an AI assistant from The Altercation Company, an Ontario company. In Augure, add an MCP server and paste the configuration above with your key.
Ollama
Ollama runs the model; it does not connect to MCP servers itself. Add AssetLab in the app you use Ollama with - Goose, Cline, VS Code, Open WebUI or any other MCP client - and choose a model that supports tool calling.
Use the core address for local models:
https://mcp.assetlab.ca/mcp?profile=core
The full server publishes 433 tools, more than most local models can hold in context. The core profile publishes 28, covering everyday records such as sites, buildings, locations, assets, work orders, work requests, PM schedules, vendors, users, asset costs and the dashboard summary. See what it leaves out in Connect Microsoft Copilot.
With a model running on your own computer, AssetLab data goes no further than that computer. A cloud model, including Ollama's own cloud models, receives what the tools return.
Any other client
The same configuration works in any client with remote MCP support. Use the plain address for the full catalogue, or the ?profile=core address when the client limits how many tools it loads. Everything a tool does is bounded by the key's scopes, whichever client calls it - see the security model.