Getting Started with Moltler¶
Prerequisites¶
- Node.js 18+ (for Skills Manager UI and MCP bridge)
- Java 21+ (for Elasticsearch plugin)
- Git
Installation¶
1. Clone the Repository¶
2. Start the Full Stack¶
This starts: - Elasticsearch with elastic-script plugin on port 9200 - Skills Manager UI on port 3000 - Loads demo data and example skills
3. Verify Installation¶
# Check Elasticsearch
curl -u elastic-admin:elastic-password http://localhost:9200
# List skills
curl -u elastic-admin:elastic-password \
-X POST http://localhost:9200/_escript \
-H "Content-Type: application/json" \
-d '{"query": "SHOW SKILLS"}'
Your First Skill¶
1. Create a Procedure¶
Procedures contain the actual logic:
CREATE PROCEDURE hello_user(name STRING)
BEGIN
PRINT 'Hello, ' || name || '!';
RETURN {"greeting": "Hello, " || name || "!", "timestamp": CURRENT_TIMESTAMP()};
END PROCEDURE;
2. Create a Skill¶
Skills wrap procedures with AI-friendly metadata:
CREATE SKILL greet_user
VERSION '1.0'
DESCRIPTION 'Greet a user by name with a friendly message'
AUTHOR 'Your Name'
TAGS ['demo', 'greeting']
(name STRING DEFAULT 'World')
RETURNS DOCUMENT
BEGIN
CALL hello_user(name);
END SKILL;
3. Invoke via MCP¶
curl -u elastic-admin:elastic-password \
-X POST http://localhost:9200/_escript/mcp \
-H "Content-Type: application/json" \
-d '{
"jsonrpc": "2.0",
"method": "tools/call",
"params": {"name": "greet_user", "arguments": {"name": "Alice"}},
"id": 1
}'
Using the Skills Manager UI¶
- Open http://localhost:3000
- Browse skills in the Skills tab
- Click a skill to see its documentation
- Click Edit to modify or Execute to run
Connecting Claude Desktop¶
- Install the MCP bridge:
- Add to Claude Desktop config (
~/Library/Application Support/Claude/claude_desktop_config.json):
{
"mcpServers": {
"moltler": {
"command": "npx",
"args": ["@moltler/mcp-bridge", "--es-url", "http://localhost:9200"]
}
}
}
-
Restart Claude Desktop
-
Ask Claude: "What tools do you have available?"
Next Steps¶
- Language Reference - Full syntax documentation
- Built-in Functions - 106 built-in functions
- MCP Integration - AI agent integration details