MCP Server

DeLive includes a standalone MCP (Model Context Protocol) server that exposes transcription data as tools and resources for AI agents like Claude Desktop, Cursor, and Claude Code.

Architecture

The MCP server is a lightweight Node.js script (mcp/delive-mcp-server.js) that:

1. Runs as a child process launched by the MCP client (e.g. Claude Desktop)
2. Uses stdio transport (standard MCP protocol)
3. Calls DeLive's REST API on http://localhost:23456 to fetch data
4. Translates MCP tool calls into HTTP requests and formats responses

MCP Client (Claude Desktop)
    ↕ stdio
DeLive MCP Server (node process)
    ↕ HTTP
DeLive App (Electron, port 23456)

Setup

Prerequisites

• DeLive running with Open API enabled (Settings > General > Open API)
• Node.js 18+
• MCP server dependencies installed: cd mcp && npm install

Environment Variables

Variable	Default	Description
DELIVE_API_URL	http://localhost:23456	DeLive REST API base URL
DELIVE_API_TOKEN	(empty)	Bearer token for authentication (set in DeLive Settings)

Client Configuration

Claude Desktop / Claude Code

Add to your Claude Desktop MCP config (claude_desktop_config.json):

{
  "mcpServers": {
    "delive": {
      "command": "node",
      "args": ["C:/path/to/DeLive/mcp/delive-mcp-server.js"],
      "env": {
        "DELIVE_API_URL": "http://localhost:23456",
        "DELIVE_API_TOKEN": "your-token-from-settings"
      }
    }
  }
}

Cursor

Add to .cursor/mcp.json (project-level) or ~/.cursor/mcp.json (global):

{
  "mcpServers": {
    "delive": {
      "command": "node",
      "args": ["C:/path/to/DeLive/mcp/delive-mcp-server.js"],
      "env": {
        "DELIVE_API_URL": "http://localhost:23456",
        "DELIVE_API_TOKEN": "your-token-from-settings"
      }
    }
  }
}

Cherry Studio

1. Open Settings > MCP Servers > Add.
2. Select stdio type.
3. Fill in the fields:
   • Command: node
   • Args: C:/path/to/DeLive/mcp/delive-mcp-server.js
   • Env: DELIVE_API_URL=http://localhost:23456, DELIVE_API_TOKEN=your-token
4. Save and enable the toggle.

OpenAI Codex CLI / Other MCP Clients

Any MCP client supporting stdio transport can launch the server directly:

DELIVE_API_URL=http://localhost:23456 \
DELIVE_API_TOKEN=your-token \
node /path/to/DeLive/mcp/delive-mcp-server.js

Tools

search_transcripts

Search sessions by keyword in title or transcript content.

Parameter	Type	Required	Default	Description
query	string	yes	—	Search keyword
limit	number	no	10	Max results

get_session

Get full session detail including transcript, AI summary, mind map, and Q&A history.

Parameter	Type	Required
sessionId	string	yes

get_session_transcript

Get transcript text only (lightweight).

Parameter	Type	Required
sessionId	string	yes

get_session_summary

Get AI summary, action items, keywords, and mind map.

Parameter	Type	Required
sessionId	string	yes

get_recording_status

Check if DeLive is currently recording and get app status.

No parameters.

list_topics

List all topic categories.

No parameters.

Resources

URI	Description
delive://sessions/recent	Most recent 10 sessions (metadata)
delive://status	Current app and recording status

Error Handling

Scenario	MCP Error Message
DeLive not running	"DeLive is not running or API is unreachable. Please start DeLive first."
API disabled	"DeLive Open API is disabled. Enable it in DeLive Settings > Open API."
Invalid token	"DeLive API token is invalid. Set the correct DELIVE_API_TOKEN environment variable."
Request timeout	10-second timeout on all HTTP calls

Dependencies

{
  "@modelcontextprotocol/sdk": "^1.29.0",
  "zod": "^4.3.6"
}