Our Mission
Give every AI agent a knowledge base
Why knowledge matters
AI agents are only as useful as the information they can access. Without the right context, even the most capable model hallucinates, gives outdated answers, or simply says “I don't know.”
Connecting your own data to an AI model should be straightforward, but today it requires stitching together dozens of libraries, managing vector databases, building parsers for every file format, and writing custom glue code for every integration. Most teams spend weeks on infrastructure before they can ask their first question.
One platform, everything included
Larkup is an open source platform that lets you build a private AI over your own files, websites, databases, and connected tools. There is no pipeline to configure before you start. Open the app, provide your API key, load content from the Data page, and start chatting.
Indexing, the retrieval server, and model selection all run behind the scenes. You focus on your data and your questions. Larkup handles the rest.
How It Works
From zero to production AI Assistant
Step 1 · Configure
Pipeline configuration
Choose your embedding model, indexing strategy, and vector store.
Project
Identifies this pipeline and names the generated server.
Embedding model
Used to embed chunks at index time and queries at runtime.
Indexing & Vector store
Choose how documents are matched at retrieval time and where vectors are stored.
Everything you need to ship AI
From loading data to deploying an API, everything is included and ready to go:
1. Load Your Data
Upload files, scrape websites, paste text, import video, or connect tools like Notion, Google Drive, and Slack. Larkup parses and indexes everything automatically.
2. Chat With Your Knowledge
Ask questions and get answers grounded in what you loaded. Responses include source citations, embedded video players, inline charts, and document previews.
3. Extend With Tools
Install community tools from the Marketplace. Add image search, document editing, video analysis, and more. Each tool runs locally and integrates directly into chat.
4. Deploy Anywhere
Launch the built in API server, grab the endpoint, and connect your agent through the TypeScript or Python SDK. Or deploy to the cloud with one click.
What we're focused on
From universal connectors to intelligent agents, everything teams need to build AI with privacy at its core:
Universal Connectors
Reliable data pipelines that sync from anywhere: Notion, Slack, Postgres, Jira, or raw files. Your knowledge base stays fresh without manual intervention.
Analytics & Insights
Built in tools to track queries, latency, and usage. See which questions your users ask, monitor performance, and identify gaps in your knowledge base.
AI Tools & Marketplace
A growing ecosystem of installable tools: image search, document editing, video analysis, and more. Every tool runs locally and plugs directly into the chat experience.
Privacy & Security
Your data never leaves your machine unless you choose to deploy. API keys are stored locally. Deploy in your own infrastructure and maintain complete data sovereignty.