Veda MCP

Veda MCP lets Claude, Cursor, Hermes, Codex, and other MCP-capable AI apps read Veda Knowledge Packs through a read-only MCP toolset.

Veda = AI-readable Knowledge Pack repository.
You ask inside your AI app; Veda supplies source-aware context through MCP.

Tools

This MCP exposes a guided Veda workflow plus read/update draft helpers:

Related MCP server: EPUB Reader MCP Server

Quick install by asking your AI

Copy this into the AI app you use for coding/automation:

Install this MCP server for me.

GitHub repository:
https://github.com/tman7162-star/veda-mcp

Use this command-based MCP server:
npx -y github:tman7162-star/veda-mcp

Environment variables:
VEDA_MCP_URL=https://veda.app/mcp
VEDA_MCP_TOKEN=<my Veda connection key here>

After installing, test it by listing my Veda Knowledge Packs.

Manual MCP config

Most MCP clients accept a config shape similar to this:

{
  "mcpServers": {
    "veda": {
      "command": "npx",
      "args": ["-y", "github:tman7162-star/veda-mcp"],
      "env": {
        "VEDA_MCP_URL": "https://veda.app/mcp",
        "VEDA_MCP_TOKEN": "veda_mcp_xxxxxxxxxxxxxxxxxxxx"
      }
    }
  }
}

For local development against a local Veda web app:

{
  "mcpServers": {
    "veda-local": {
      "command": "node",
      "args": ["C:/path/to/veda-mcp/src/index.js"],
      "env": {
        "VEDA_MCP_URL": "http://127.0.0.1:8795/mcp",
        "VEDA_MCP_TOKEN": "veda_mcp_xxxxxxxxxxxxxxxxxxxx"
      }
    }
  }
}

Environment variables

Security notes

• Treat VEDA_MCP_TOKEN connection key like a password.

• Start with read-only scopes: pack:read and pack:context.

• If the token leaks, revoke/regenerate it in Veda.

• Do not paste your token into public GitHub issues, screenshots, or commits.

Development

npm install
npm run check
npm run smoke:list-tools

smoke:list-tools verifies that the local stdio MCP server can start and list its tool definitions. It does not require a valid Veda token because it does not call Veda tools.

License

MIT