graphlit

This MCP server acts as the bridge between MCP clients and the Graphlit platform, enabling ingestion from various sources (Slack, Discord, websites, email, Jira, GitHub, and more) into a Graphlit project and exposing a suite of MCP tools for searching, retrieving, and manipulating knowledge. The server supports a range of capabilities including retrieval (querying contents, feeds, collections, conversations; retrieving sources and similar images), RAG support (prompting and conversations with LLMs), extraction (structured JSON from text), publishing (audio or image output), ingestion (files, web pages, messages, emails, issues, text, memory), data connectors (notable apps like Notion, Google Drive, Jira, Slack, GitHub, Outlook, Drive, Box, Redis, etc.), web tools (web crawling and search), notifications (via Slack, email, webhooks, Twitter), and operational tasks (configuring projects, managing collections and contents, and status checks). You can use these tools to ingest data, organize it into collections, and query it effectively from MCP clients like Cursor, Windsurf, Goose, or Cline. In short, it enables end-to-end data ingestion to a Graphlit knowledge base and provides robust MCP-based access for applications that need to search and reason over that data.

Prerequisites:

• Node.js (recommended v18+)
• npm or pnpm
• Access to a Graphlit Platform account with API settings (for environment variables)

Installation steps:

1. Install Node.js if you don’t have it already. Visit https://nodejs.org/ and follow the installer for your OS.

2. Install and run the MCP server using NPX (no global install required):

npx -y graphlit-mcp-server

3. Configure environment variables (example for development):

export GRAPHLIT_ORGANIZATION_ID=<your-organization-id>
export GRAPHLIT_ENVIRONMENT_ID=<your-environment-id>
export GRAPHLIT_JWT_SECRET=<your-jwt-secret>

4. If you prefer a VS Code-based setup, you can add the following JSON block to your settings or workspace file as shown in the README (the example uses the same environment variable placeholders):

{
  "mcp": {
    "inputs": [
      {
        "type": "promptString",
        "id": "organization_id",
        "description": "Graphlit Organization ID",
        "password": true
      },
      {
        "type": "promptString",
        "id": "environment_id",
        "description": "Graphlit Environment ID",
        "password": true
      },
      {
        "type": "promptString",
        "id": "jwt_secret",
        "description": "Graphlit JWT Secret",
        "password": true
      }
    ],
    "servers": {
      "graphlit": {
        "command": "npx",
        "args": ["-y", "graphlit-mcp-server"],
        "env": {
          "GRAPHLIT_ORGANIZATION_ID": "${input:organization_id}",
          "GRAPHLIT_ENVIRONMENT_ID": "${input:environment_id}",
          "GRAPHLIT_JWT_SECRET": "${input:jwt_secret}"
        }
      }
    }
  }
}

5. Alternative: Windsurf or other IDEs often provide a one-click install flow that runs npx -y graphlit-mcp-server with your environment variables filled in via prompts. See the README for the specific install buttons and example URLs.

6. Verify the server is running by checking logs and ensuring the MCP endpoints respond to client requests as documented by Graphlit.

Tips and considerations:

• Keep GRAPHLIT_JWT_SECRET and other credentials secure; never commit them to public repos.
• The mcp configuration uses the graphlit-mcp-server package name; you can adjust the server key (e.g., graphlit-mcp-server) to match your workspace naming conventions.
• If you’re using Windsurf, the example shows how to pass organization, environment, and JWT via environment variables; adapt to your deployment environment accordingly.
• The Tools list in the README indicates a broad set of data ingestion, retrieval, and publishing capabilities; plan collections and data connectors to fit your use case (e.g., ingest Notion pages or Google Drive documents, then query with retrieval and RAG tools).
• For local development, you can run npx -y graphlit-mcp-server and then connect MCP clients to the server’s endpoints. Check the Graphlit portal for API keys and environment IDs.
• If you encounter authentication errors, double-check that GRAPHLIT_ORGANIZATION_ID, GRAPHLIT_ENVIRONMENT_ID, and GRAPHLIT_JWT_SECRET are correctly set and that the Graphlit account has API access enabled.