Graphlit MCP Server
The official Graphlit MCP server: an agent collects material from Slack, GitHub, mail and websites into a knowledge base and searches it
- 378 ★ on the repository
- Last commit
- MIT
Broad access
Why this level
- The JWT secret opens the Graphlit project holding material from mail, chats and drives
- The set includes deleting collections, feeds and contents and sending notifications to Slack, email and X
- Data goes to the Graphlit cloud and, when publishing, to ElevenLabs and OpenAI
Install
- Works in
In your terminal, with SkillFoxx CLI
npx skillfoxx add mcp/graphlit-mcp-serverDetects the agents on your machine, shows the access level and pins the version.
Other ways to install
Access and checks
How we reviewBroad accessThe tool can change data, run commands or handle secrets.
- The JWT secret opens the Graphlit project holding material from mail, chats and drives
- The set includes deleting collections, feeds and contents and sending notifications to Slack, email and X
- Data goes to the Graphlit cloud and, when publishing, to ElevenLabs and OpenAI
This is third-party code. Review the repository files before installing.
What it does
Graphlit MCP Server connects an MCP client such as Cursor, Windsurf, Goose or Cline to the Graphlit platform. An agent ingests material from Slack, Discord, websites, Google Drive, mail, Jira, Linear and GitHub into a Graphlit project, then searches and retrieves what it needs: the project acts as a RAG-ready knowledge base. PDF, DOCX and PPTX documents and web pages become Markdown on ingestion, and audio and video are transcribed. The tools are split into groups. Retrieval covers queries over contents, collections, feeds and conversations, relevant-source retrieval, similar-image search and image description. RAG offers a conversation call to a model, and a separate tool extracts structured JSON from text. Ingestion handles files, web pages, messages, posts, emails, issues, text and short-term memory. Connectors cover Outlook, Gmail, Notion, Reddit, Linear, Jira, GitHub, Google Drive, OneDrive, SharePoint, Dropbox, Box, Slack, Teams, Discord, X and podcasts over RSS. Web crawling, web search, site mapping and page screenshots are built in, so separate Firecrawl or Exa are not needed. There is publishing to audio through ElevenLabs and to images through OpenAI, notifications to Slack, email, webhook and X, plus collection operations and deletion.
Who it is for. For developers and product teams who want to give an agent in Cursor or Cline a shared knowledge base built from work chats, tickets, documents and sites.
Good fit when
- You want to gather Slack discussions, Jira and Linear tickets and Google Drive documents into one base and then ask questions over them
- An agent should analyze a competitor's site: crawl pages, get Markdown and find the relevant parts
- You need to transcribe a video or podcast and search the transcript
- A long text must become structured JSON following a given schema
- You want a Slack or email notification when a feed finishes ingesting
Not a fit when
- You have no Graphlit account: without the environment and organization IDs and the JWT secret the server cannot connect
- Data must not go to a third-party cloud: materials are loaded into a project on the Graphlit platform
- You only need web search or page reading: a lighter server with no connectors or account fits better
- The agent must not get delete and broadcast rights: the set includes deleting collections, feeds and contents, and sending notifications
Example request
Ingest the documentation pages at this address into the Graphlit project, wait for processing to finish and find the sections about authentication.MCP
- Transport
- stdio
- Authentication
- API key
Environment variables
GRAPHLIT_ORGANIZATION_IDrequired- Organization ID on the Graphlit platform.
GRAPHLIT_ENVIRONMENT_IDrequired- Project environment ID.
GRAPHLIT_JWT_SECRETrequired, secret- Secret for signing the JWT, opens access to the project.
Limitations
You need Node.js 18 or newer and an active account on the Graphlit platform. The connection is configured with three variables: GRAPHLIT_ENVIRONMENT_ID, GRAPHLIT_ORGANIZATION_ID and GRAPHLIT_JWT_SECRET. All values come from the API settings dashboard on the Graphlit portal. The JWT secret opens access to the project, so in VS Code it is entered through a protected field, while in other clients it sits in the config file.
Material is stored and processed on the Graphlit side, and publishing to audio and images uses the third-party services ElevenLabs and OpenAI. Connectors to mail, drives and messengers need separate authorization in those services. Plan terms and platform limits are not given in the README, see the vendor's site.
The last commit dates from January 2026. Availability of the Graphlit platform from Russia was not checked when this card was prepared.
How to disable. Remove the graphlit block from the MCP client config and restart the client. Then revoke the JWT secret in the Graphlit dashboard if it could have leaked, and delete unneeded feeds and project contents.
FAQ
How do I connect Graphlit MCP Server to Cursor?
Add a server to `mcp.json` with the `npx` command and the arguments `-y`, `graphlit-mcp-server`, and set GRAPHLIT_ORGANIZATION_ID, GRAPHLIT_ENVIRONMENT_ID and GRAPHLIT_JWT_SECRET in env. The values come from the Graphlit dashboard. The block is the same for Windsurf and Cline.
What do I need to use Graphlit MCP Server?
Node.js 18 or newer and an account on the Graphlit portal with access to API settings. Without the three environment variables the server will not start. Connectors also need authorization in the relevant services.
Which sources can Graphlit MCP Server ingest?
Files, web pages, messages, posts, emails, issues and text, and through connectors Slack, Teams, Discord, Gmail, Outlook, Notion, Jira, Linear, GitHub, Google Drive, OneDrive, SharePoint, Dropbox, Box, Reddit, X and RSS podcasts. Documents become Markdown and audio and video are transcribed.
Do I need a separate service for crawling sites and web search?
No. Web crawling, web search including podcast search, site mapping and page screenshots are built into the server, so Firecrawl or Exa need not be connected.
How does Graphlit MCP Server differ from Firecrawl MCP Server?
Firecrawl focuses on web search and getting pages as Markdown. Graphlit also keeps material in a project as a knowledge base, pulls data from connectors to messengers, mail and drives and can hold conversations with a model over what is stored.
Discussion
No questions or reviews yet.
Related
MCP serversMCP server for the Flexible GraphRAG platform: ingest from 14 data sources, hybrid search and AI answers over a knowledge graph
A CLI and MCP server that extracts structured knowledge from documents: graphs, hypergraphs, and fact search with sources
A local MCP server that turns a folder of documents into a knowledge base with hybrid search for Claude Code and Cursor
A self-hosted RAG app with a web interface and a built-in MCP server: documents, Qdrant and answers with sources.
A knowledge graph with an MCP server: the agent searches concepts by meaning and sees how well each is supported.
InfraNodus MCP server turns any text into a knowledge graph and shows topic clusters, the gaps between them and research questions
Comparison with similar tools
Cards that share the main search query, and how they differ.
| Tool | Official | Authentication | Repository stars | Main difference |
|---|---|---|---|---|
| Graphlit MCP Server (this card) | yes | API key | 378 | |
| Firecrawl MCP Server | yes | API key | 7,765 | the official Firecrawl server for web search and converting pages to Markdown |
| Basic Memory | no | not required | 4,495 | long-term agent memory in local Markdown files with a link graph |
README in short
The README begins by describing the server's purpose with links to the blog, a video and the community. Then come the tool list by group, resources, prerequisites, environment variables and install methods for VS Code, Windsurf, Cline and Cursor. Smithery is mentioned separately.