# Weavestream as an MCP Server

Weavestream can expose your data to any AI assistant that supports the Model Context Protocol (MCP) — including Claude Desktop, Gemini, and others. Once enabled, those AI tools can search, filter, and query your Weavestream data directly from their own interfaces.

# How It Works

When the MCP server is running, it starts a local HTTP server on your Mac that listens for requests from connected AI clients. The server is only accessible from your Mac (127.0.0.1) — it cannot be reached from the internet or other devices on your network.

Connected AI clients can:

Discover your sources, endpoints, RSS feeds, and Apple Notes
Search across your items using keywords
Filter items by field conditions
Count items and get breakdowns by source or status
Read detailed metadata about your data sources

# Enabling the MCP Server

Open Settings (Cmd+Comma) and go to the MCP Server tab
Toggle on Enable MCP Server
The server starts immediately — you'll see a green indicator and the address it's running on

# Configuration

# Port

The server runs on port 9735 by default. You can change this to any port between 1024 and 65535. If another app is already using the default port, change it here.

# Authentication Token

Every request to the MCP server requires a Bearer token. Weavestream generates this automatically. You'll see the token masked in the settings (showing only the first few characters).

Copy — Copies the full token to your clipboard for pasting into client config files
Regenerate — Creates a new token and invalidates any existing client configurations. You'll need to update all connected clients after regenerating.

# Status

When the server is running, the settings pane shows:

The address it's listening on (e.g., Running on 127.0.0.1:9735)
The name of the last AI client that connected
A count of how many requests have been served in the current session

# Connecting an AI Client

# Claude Desktop

Add the following to your claude_desktop_config.json file (Weavestream shows this snippet with your actual port and token pre-filled):

{
  "mcpServers": {
    "weavestream": {
      "command": "npx",
      "args": [
        "-y",
        "mcp-remote",
        "http://localhost:9735/mcp",
        "--header",
        "Authorization: Bearer YOUR_TOKEN"
      ]
    }
  }
}

# Google Gemini

Add the following to your .gemini/settings.json file:

{
  "mcpServers": {
    "weavestream": {
      "url": "http://localhost:9735/mcp",
      "headers": {
        "Authorization": "Bearer YOUR_TOKEN"
      }
    }
  }
}

The settings pane shows both snippets with your actual port and token already filled in — just copy and paste.

# What the AI Can Access

# Tools

Connected AI clients have access to the same tools Weavestream's built-in agent uses:

Tool	What It Does
`search_items`	Full-text search across item titles, summaries, and raw data
`filter_items`	Filter items by field conditions
`count_items`	Count items with optional grouping
`get_item_details`	Fetch full details for a specific item
`list_sources`	Discover available sources, endpoints, and fields
`get_field_values`	Get distinct values for a specific field

# Resources

The MCP server also exposes browsable resources, giving AI clients a structured overview of your data:

All sources (weavestream://sources) — Overview of every source, endpoint, RSS feed, and Apple Notes with item counts
Individual sources (weavestream://sources/{id}) — Details for a specific API source
Endpoints (weavestream://endpoints/{id}) — Fields and item counts for a specific endpoint
RSS categories and feeds — Counts and field info for your feeds
Apple Notes (weavestream://apple-notes) — Item counts and available fields

# Privacy & Security

The server only accepts connections from your Mac (127.0.0.1) — remote access is not possible
Every request is authenticated with the Bearer token
Your Weavestream credentials (API keys, OAuth tokens) are never exposed to connected clients — only item data retrieved through tools
Stopping the server (toggle off) immediately closes all connections