Labsco
MCP SERVER · OFFICIAL PROJECT

Graphlit

by graphlit ↗

Pull Slack, Discord, Teams, Jira, Linear, GitHub, Notion, email, cloud drives and crawled sites into one knowledge base, then ask questions across all of it in the same session.

Vector Stores & RAG Retrieval✓ Official source
Summary
One place to ask a question that currently needs four tabs and a Jira search.

The reach is the point: the same knowledge base holds a Slack thread, the Linear issue it argued about, the PDF someone dropped in Dropbox and the crawled docs page that settles it, and one retrieval call spans all four. That reach is bought one credential at a time, connector by connector, and a connector whose environment variable is missing is a tool that will refuse. Ingestion runs asynchronously and returns an identifier rather than content, so a workflow built on it has to poll isContentDone or isFeedDone before it queries.

What it is

An ingestion and retrieval front end for the Graphlit platform: 71 tools that read content out of chat platforms, issue trackers, mailboxes, cloud drives and the open web, store it, and answer questions over what has been stored.

What you get
  • Ingestion from cloud file stores — Google Drive, OneDrive, Dropbox, Box, SharePoint libraries and GitHub repositories — with a read limit per run and folder identifiers you can enumerate first
  • Ingestion from conversations: Slack and Discord channels, Microsoft Teams channels, Reddit subreddits, Twitter/X posts by user or by search, and Google or Microsoft email
  • Issue history from Linear projects, GitHub repositories and Jira, and pages from a Notion database
  • Web intake three ways: crawl a site through its sitemap, enumerate URLs without ingesting them, or search the web and podcasts by query
  • Retrieval over everything stored — vector search for sources, image similarity search, and a conversation that keeps its history between turns
  • Extraction of JSON from text against a schema you supply, and vision descriptions of an image URL or a stored image
  • Feeds that re-check for new content every 15 minutes when recurring is set, with isFeedDone and isContentDone to tell you when an async run has finished
  • Outbound notifications to a Slack channel, an email address, a webhook or Twitter/X, plus generated audio and images stored as content
Requirements

A Graphlit project, identified by GRAPHLIT_ORGANIZATION_ID, GRAPHLIT_ENVIRONMENT_ID and GRAPHLIT_JWT_SECRET. Each connector needs its own credential on top of that — SLACK_BOT_TOKEN for Slack, DISCORD_BOT_TOKEN for Discord, NOTION_API_KEY for Notion, LINEAR_API_KEY for Linear, GITHUB_PERSONAL_ACCESS_TOKEN for GitHub, JIRA_EMAIL with JIRA_TOKEN for Jira, TWITTER_TOKEN for Twitter/X, and GOOGLE_EMAIL_CLIENT_ID, GOOGLE_EMAIL_CLIENT_SECRET and GOOGLE_EMAIL_REFRESH_TOKEN for Gmail. MIT licensed.

Setup effort

One command plus a key — npx -y graphlit-mcp-server, then supply credentials