graphlit

Overview

The graphlit MCP server is a Model Context Protocol (MCP) server that enables AI assistants and agents to interact directly with Graphlit, a platform for ingesting, processing, organizing, and querying unstructured content using knowledge graphs and embeddings. It allows AI-driven workflows to load documents, media, and web content into Graphlit and then reason over that content using structured queries — without building custom ingestion or retrieval pipelines.

This server is especially useful for knowledge management, content intelligence, retrieval-augmented generation (RAG), and research workflows that span large, heterogeneous content collections.

Transport

stdio

Tools

• addContentsToCollection
• askGraphlit
• configureProject
• createCollection
• deleteCollection
• deleteCollections
• deleteContent
• deleteContents
• deleteConversation
• deleteConversations
• deleteFeed
• deleteFeeds
• describeImageContent
• describeImageUrl
• extractText
• ingestBoxFiles
• ingestDiscordMessages
• ingestDropboxFiles
• ingestFile
• ingestGitHubFiles
• ingestGitHubIssues
• ingestGoogleDriveFiles
• ingestGoogleEmail
• ingestJiraIssues
• ingestLinearIssues
• ingestMemory
• ingestMicrosoftEmail
• ingestMicrosoftTeamsMessages
• ingestNotionPages
• ingestOneDriveFiles
• ingestRSS
• ingestRedditPosts
• ingestSharePointFiles
• ingestSlackMessages
• ingestText
• ingestTwitterPosts
• ingestTwitterSearch
• ingestUrl
• isContentDone
• isFeedDone
• listBoxFolders
• listDiscordChannels
• listDiscordGuilds
• listDropboxFolders
• listGoogleCalendars
• listLinearProjects
• listMicrosoftCalendars
• listNotionDatabases
• listNotionPages
• listSharePointFolders
• listSharePointLibraries
• listSlackChannels
• promptConversation
• publishAudio
• publishImage
• queryCollections
• queryContents
• queryConversations
• queryFeeds
• queryProjectUsage
• removeContentsFromCollection
• retrieveImages
• retrieveSources
• screenshotPage
• sendEmailNotification
• sendSlackNotification
• sendTwitterNotification
• sendWebHookNotification
• webCrawl
• webMap
• webSearch

Key Capabilities

• Unified content ingestion — Load files, web pages, feeds, and other unstructured sources into a single system.
• Knowledge graph construction — Extract entities, relationships, and metadata for structured reasoning.
• Semantic retrieval — Query content using embeddings and natural-language intent.
• Metadata- and filter-aware queries — Combine semantic relevance with structured constraints.
• RAG-ready workflows — Provide high-quality, curated context to downstream language models.

How It Works

The graphlit MCP server runs as an MCP service that connects AI clients to the Graphlit platform using secure, scoped credentials. Once configured, the server exposes Graphlit’s content ingestion and query capabilities over the MCP protocol.

AI assistants can request that new content be ingested into Graphlit or retrieve relevant content from existing collections as part of a broader reasoning workflow. The server handles communication with Graphlit, content normalization, and response formatting, returning structured results that models can reason over directly.

By abstracting away ingestion pipelines, embedding management, and graph construction, the server allows AI workflows to focus on using knowledge, not maintaining infrastructure. This makes it possible to build assistants that answer questions, synthesize insights, or generate outputs grounded in large, evolving content libraries.