Getting Started with laurus-mcp

Prerequisites

• The laurus CLI binary installed (cargo install laurus-cli)
• A running laurus-server instance (see laurus-server getting started)
• An AI client that supports MCP (Claude Desktop, Claude Code, etc.)

Configuration

Step 1: Start laurus-server

laurus serve --grpc-port 50051

Step 2: Configure the MCP client

Claude Code

Use the CLI command (recommended):

claude mcp add laurus laurus mcp --endpoint http://localhost:50051

Or edit ~/.claude/settings.json directly:

{
  "mcpServers": {
    "laurus": {
      "command": "laurus",
      "args": ["mcp", "--endpoint", "http://localhost:50051"]
    }
  }
}

Claude Desktop

Edit ~/Library/Application Support/Claude/claude_desktop_config.json (macOS) or %APPDATA%\Claude\claude_desktop_config.json (Windows):

{
  "mcpServers": {
    "laurus": {
      "command": "laurus",
      "args": ["mcp", "--endpoint", "http://localhost:50051"]
    }
  }
}

Usage Workflows

Workflow 1: Pre-created index

Create the index using the CLI first, then use the MCP server to query it:

# Step 1: Create a schema file
cat > schema.toml << 'EOF'
[fields.title]
Text = { indexed = true, stored = true }

[fields.body]
Text = { indexed = true, stored = true }
EOF

# Step 2: Start the server and create the index
laurus serve --grpc-port 50051 &
laurus create index --schema schema.toml

# Step 3: Start the MCP server (done automatically by Claude)
laurus mcp --endpoint http://localhost:50051

Workflow 2: AI-driven index creation

Start the MCP server without a pre-created index, then ask the AI to create one:

# Start laurus-server (no index required)
laurus serve --grpc-port 50051

Then ask Claude:

“Create a search index for blog posts. I need to search by title and body text, and I want to store the author and publication date.”

Claude will call the connect tool (if not already connected), then design the schema and call create_index automatically.

Workflow 3: Connect at runtime

Start the MCP server without specifying an endpoint:

laurus mcp

Then ask Claude to connect:

“Connect to the laurus server at http://localhost:50051”

Claude will call connect(endpoint: "http://localhost:50051") before using other tools.

Lifecycle

laurus-server starts (separate process)
  └─ listens on gRPC port 50051

Claude starts
  └─ spawns: laurus mcp --endpoint `http://localhost:50051`
       └─ enters stdio event loop
            ├─ receives tool calls via stdin
            ├─ proxies calls to laurus-server via gRPC
            └─ sends results via stdout
Claude exits
  └─ laurus-mcp process terminates
  └─ laurus-server continues running