Graphlit MCP Server – Knowledge Ingestion and RAG
Graphlit provides an official open-source MCP server that connects MCP clients to the Graphlit Platform. It lets AI agents ingest files, web pages, messages, emails, issues, and connector data into a Graphlit project, then search, retrieve, and use that knowledge in RAG workflows.
Overview
Graphlit MCP Server is the official Model Context Protocol server for the Graphlit Platform. It gives MCP-compatible clients a structured way to work with a Graphlit project as a searchable, RAG-ready knowledge base. The server is useful when an agent needs current project context from documents, web pages, developer tools, product-management tools, communications, and other connected sources rather than relying only on model memory.
What the MCP server enables
The server exposes Graphlit tools and resources for retrieval, ingestion, RAG, extraction, publishing, web operations, notifications, project operations, and connector enumeration. Document and HTML content can be extracted to Markdown during ingestion, while audio and video files can be transcribed. Agents can query contents, collections, feeds, and conversations; retrieve relevant sources; retrieve similar images; visually describe images; and prompt LLM conversations over Graphlit-managed context.
It also supports ingestion of files, web pages, messages, posts, emails, issues, text, and short-term memory. Officially documented connectors include services such as Microsoft Outlook, Google Mail, Notion, Reddit, Linear, Jira, GitHub Issues, Google Drive, OneDrive, SharePoint, Dropbox, Box, GitHub, Slack, Microsoft Teams, Discord, Twitter/X, and podcast RSS feeds. Web crawling, web search, web mapping, and page screenshots are available as MCP tools.
When to use it
Use this MCP server when an agent needs to build or query a project knowledge base across many content sources. Practical workflows include asking an IDE assistant to search Slack and GitHub context, retrieving related documents before drafting a response, ingesting a web page for later analysis, checking whether a feed or content item has completed processing, or organizing retrieved content into Graphlit collections.
Connection and authentication
Graphlit documents a local stdio MCP server launched with npx -y graphlit-mcp-server. Authentication and project selection require GRAPHLIT_ORGANIZATION_ID, GRAPHLIT_ENVIRONMENT_ID, and GRAPHLIT_JWT_SECRET, which are found in the Graphlit Platform API settings dashboard. Optional connector credentials can be added for services such as Slack, Discord, Google Email, Linear, GitHub, Jira, and Notion when those connector tools are needed.
Key considerations
The documented installation path is a local Node.js-based stdio server and requires Node.js 18.x or higher. Keep the Graphlit JWT secret and connector credentials out of shared configuration files. Grant only the connector tokens an agent needs, because the server can ingest, query, create collections, delete collections, delete feeds, delete content, and delete conversations when configured with suitable access.
Supported Transports
stdio
Command: npx
Args:
-ygraphlit-mcp-server
Frequently Asked Questions
- When should an AI agent use the Graphlit MCP Server?
- Use it when an agent needs to ingest, search, retrieve, and reason over a Graphlit project knowledge base that can include documents, web pages, messages, emails, issues, source-tool data, and short-term memory.
- What does the Graphlit MCP Server add to an AI agent's capabilities?
- It adds live access to Graphlit-managed retrieval, RAG conversations, structured extraction, ingestion, web crawling, web search, web mapping, screenshots, content organization, and connector enumeration tools.
- What can the AI agent access or manage through this server?
- The documented server exposes Graphlit resources such as projects, contents, feeds, collections, workflows, conversations, and specifications, plus tools to create collections, add or remove content from collections, delete collections, feeds, content, and conversations, and check whether feed or content processing is done.
- How is authentication configured for the Graphlit MCP Server?
- The server uses Graphlit project credentials supplied as environment variables: GRAPHLIT_ORGANIZATION_ID, GRAPHLIT_ENVIRONMENT_ID, and GRAPHLIT_JWT_SECRET. These values come from the Graphlit Platform API settings dashboard and should be stored as secrets rather than committed to shared configuration.
- Which transport should be used for the Graphlit MCP Server?
- Use the documented local stdio transport by launching npx with the graphlit-mcp-server package. The official setup examples for VS Code, Windsurf, Cline, Cursor, and generic MCP clients all use this local npx stdio command.