Ori runs a local MCP (Model Context Protocol) server on localhost:4141. This lets other tools use Ori’s capabilities — memory, tools, and intelligence.
What this enables
| Tool | How it uses Ori |
|---|
| Cursor | Ask Ori questions directly from your editor |
| VS Code | Use Ori as an AI backend with memory |
| CLI scripts | Call Ori from shell scripts and automation |
| Custom tools | Any MCP-compatible client can connect |
The MCP server exposes 7 tools:
| Tool | Description |
|---|
ask_ori | Send a natural language query to Ori |
read | Read files or directories |
write | Create or modify files |
run | Execute shell commands |
search | Search the web |
crawl | Extract content from URLs |
system | Get system information |
Connecting from Cursor
Add to your Cursor MCP config:
{
"mcpServers": {
"ori": {
"url": "http://127.0.0.1:4141/mcp"
}
}
}
Connecting from VS Code
Add to your VS Code settings:
{
"mcp": {
"servers": {
"ori": {
"url": "http://127.0.0.1:4141/mcp"
}
}
}
}
Connecting via CLI
# Health check
curl http://127.0.0.1:4141/
# MCP endpoint (POST)
curl -X POST http://127.0.0.1:4141/mcp \
-H "Content-Type: application/json" \
-d '{"method": "tools/call", "params": {"name": "ask_ori", "arguments": {"query": "What time zone am I in?"}}}'
The MCP server only accepts connections from localhost. It’s not exposed to the network — only tools running on your machine can connect.