Jarvis

You can deploy and run the Jarvis MCP Server to connect Claude Desktop with n8n and Supabase for autonomous Jarvis agent tuning. It exposes tools for semantic search, memory, workflow orchestration, task creation, and agent health checks, and is designed for production-ready, zero-downtime deployment.

How to use

Connect your MCP client to the Jarvis MCP Server using the provided stdio configuration. You will run a local command that invokes curl to call the MCP endpoint, with an environment variable that points to your server URL. Once connected, you can access the available tools to execute semantic search, save context, trigger workflows, create tasks across platforms, and monitor agent status.

How to install

Prerequisites: ensure you have a modern runtime environment available (the server is designed for use in Replit deployments and similar environments). Follow these concrete steps to set up the Jarvis MCP Server for first run.

Step 1: In Replit, establish a GitHub connection in Settings.

Step 2: Set Secrets according to the provided environment guidance (see .env.example for the required keys and values). Then save your changes.

Step 3: Click Run to enable Auto-Deploy and start serving MCP calls. The server will expose its endpoints and be ready to accept tool calls from Claude Desktop.

Configuration and runtime example

{
  "mcpServers": {
    "jarvis": {
      "command": "curl",
      "args": ["-X", "POST", "https://YOUR-REPLIT-URL/mcp/tools/call"],
      "env": {
        "MCP_SERVER_URL": "https://YOUR-REPLIT-URL"
      }
    }
  }
}

Other important notes

The server includes a production-ready HTTP MCP setup designed for Claude Desktop integration, with features such as semantic search via pgvector, persistent memory of interactions, workflow orchestration through n8n, cross-platform task creation, and continuous agent health monitoring.

Available tools

conversation_search

Semantic search in Jarvis Knowledge Base using pgvector for RAG-based retrieval.

save_context

Persist Claude-Jarvis interactions for future retrieval and memory.

trigger_workflow

Start n8n workflows directly from the MCP server to automate tasks.

create_task

Create tasks across multiple platforms such as Monday.com, GitHub, and Slack.

get_agent_status

Monitor agent health and status checks to ensure production readiness.